CN116846978A - Resource scheduling method, application identification method and related equipment of cloud computing system - Google Patents

Abstract

Embodiments of this application disclose resource scheduling methods, application identification methods and related equipment of cloud computing systems, which are used to improve resource utilization and reduce interference between applications in the computer system. The resource scheduling method is applied to the resource scheduler, and the method includes: obtaining the remaining interference tolerance of each LC type application in at least one delay-sensitive LC type application included in the computer system; obtaining the remaining interference tolerance from multiple LC type applications Smallest first LC type application. If the remaining interference tolerance of the first LC type application is less than the lower limit of the tolerance, the first isolation area resources of the first LC type application are increased. If the remaining interference tolerance of the first LC type application is greater than the tolerance upper limit, the second isolation area resources of the second LC type application among the multiple LC type applications are transferred to the resource sharing area.

Description

Resource scheduling method, application identification method and related equipment of cloud computing system

Technical Field

The embodiments of the present application relate to the computer field, and in particular to a resource scheduling method, an application identification method, and related devices of a cloud computing system.

Background Art

With the continuous development of the Internet, the scale of data centers as information infrastructure is also growing. However, the resource utilization rate in most data centers is still very low. In order to reduce costs, multiple applications can be run simultaneously to share the underlying resources to improve resource utilization. Although the co-location of applications can effectively improve resource utilization, applications deployed on the same physical machine will compete for shared resources, resulting in frequent interference between applications.

In a resource scheduling method, each latency critical (LC) application has fixed isolation zone resources, and non-isolated shared resources are processed based on reinforcement learning to determine a resource allocation scheme that meets the service target quality of the LC application.

In this method, since each LC type application has fixed isolation area resources, when the application requires fewer resources, the fixed isolation area resources occupied by the application may be much greater than the actual demand, resulting in low resource utilization.

Summary of the invention

The embodiment of the present application provides a resource scheduling method, an application identification method and related devices of a cloud computing system. In the resource scheduling method, different resource scheduling modes are determined according to different values of the residual interference tolerance of each LC type application in the computer system, so as to increase the isolation area resources of the LC type application, or reduce the isolation area resources of the LC type application to increase the shared area resources of the computer system, thereby realizing the flexible allocation of isolation area resources and shared area resources and improving the utilization rate of resources. At the same time, by scheduling isolation area resources and shared area resources, the interference between applications is reduced and the system performance is also improved.

A first aspect of an embodiment of the present application provides a resource scheduling method for a cloud computing system, the method being applied to a resource scheduler, the method comprising:

There are multiple LC type applications running in the computer system, and the resource scheduler can calculate or receive the remaining interference tolerance of each LC type application from the computing device. The remaining interference tolerance can reflect the degree of interference of each LC type application. The smaller the remaining interference tolerance, the greater the degree of interference of the LC type application. Afterwards, the resource scheduler selects the first LC type application with the smallest remaining interference tolerance from the multiple LC type applications. And compare the relationship between the remaining interference tolerance of the first LC type application and the tolerance lower limit to determine whether it is necessary to adjust the isolation area resources of the first LC type application. If the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, it means that the first LC type application is seriously interfered with and the resources are tight, then the first isolation area resources of the first LC type application are increased. If the remaining interference tolerance of the first LC type application is greater than the tolerance upper limit, and because the first LC type application has the smallest remaining interference tolerance among all LC type applications, it can be considered that all LC type applications are not interfered with or the interference is small, and the second isolation area resources of the second LC type application in the multiple LC type applications can be transferred to the resource sharing area. The residual interference tolerance of the second LC type application is greater than the residual interference tolerance of the first LC type application. Optionally, the residual interference tolerance of the second LC type application may be the largest among the multiple LC type applications.

It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

According to the different values of the residual interference tolerance of each LC type application in the computer system, different resource scheduling methods are determined to increase the isolated area resources of the LC type application, or reduce the isolated area resources of the LC type application to increase the shared area resources of the computer system, thereby realizing the flexible allocation of isolated area resources and shared area resources and improving resource utilization. At the same time, by scheduling isolated area resources and shared area resources, the interference between applications is reduced and the system performance is improved.

In some optional embodiments of the first aspect, the resource scheduler may also obtain the current system entropy of the computer system, and the current system entropy is used to indicate the interference degree between applications in the current computer system. The larger the current system entropy is, the greater the interference degree between applications in the current indicated computer system is, and the worse the performance of the computer system is. The resource scheduler determines the area for resource scheduling by comparing the size relationship between the current system entropy and the system entropy threshold, and the size relationship between the remaining interference tolerance of the first LC type application and the tolerance lower limit and the tolerance upper limit. Specifically, if the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, then it can be considered that the first LC type application is seriously interfered with, resources are tight, and the first isolation area resources of the first LC type application need to be increased. If the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is greater than the tolerance upper limit, then it can be considered that the poor performance of the current system is due to the large number of isolation area resources of the LC type application, so the resource scheduler can transfer the second isolation area resources of the second LC type application to the resource sharing area.

In the embodiment of the present application, the remaining tolerance and system entropy are combined to determine whether to perform resource scheduling, and the conditions for resource scheduling are more strictly limited, thereby avoiding resource scheduling under unnecessary circumstances and saving computing resources.

In some optional embodiments of the first aspect, there are multiple ways for the resource caller to increase the first isolation area resources of the first LC type application. Optionally, if there is a third LC type application in the computer system whose residual interference tolerance is greater than the tolerance upper limit and has a strippable third isolation area resource, the resource scheduler will transfer the resources in the third isolation area resources to the first isolation area corresponding to the first LC type application to increase the first isolation area resources. Optionally, if there is no such third LC type application in the computer system, the resource scheduler will transfer the resources of the resource sharing area to the first isolation area to increase the first isolation area resources. Among them, the absence of such a third LC type application in the computer system includes that the residual interference tolerance of all LC type applications in the computer system is not greater than the tolerance upper limit, or, all LC type applications in the computer system do not have strippable isolation area resources, or, the LC type applications in the computer system whose residual interference tolerance is greater than the tolerance upper limit do not have strippable isolation area resources, and the specifics are not limited here.

In the embodiment of the present application, when increasing the first isolation area resources of the first LC type application, priority is given to transferring resources from the isolation area resources of other LC type applications, so as to minimize the adverse effects on system performance and improve the practicality of the technical solution. At the same time, for different situations, there are different ways to increase the first isolation area resources, which can adapt to different scenarios and improve the flexibility and adaptability of the technical solution of the present application.

In some optional embodiments of the first aspect, when the computer system includes only LC type applications, the resource scheduler can obtain the amount of interference that each LC type application can tolerate and the amount of interference that each LC type application actually receives, and determine the current system entropy based on the amount of interference that each LC type application can tolerate and the amount of interference that each LC type application actually receives.

In some optional embodiments of the first aspect, when a computer system includes an LC type application and at least one best effort (BE) type application, the resource scheduler may obtain a first application identifier and a second application identifier from an application distinguisher or from a user to distinguish between the LC type application and the BE type application. The first application identifier is used to indicate the LC type application, and the second application identifier is used to indicate the BE type application. The resource scheduler can determine the LC type application in the computer system according to the first application identifier. According to the second application identifier, the BE type application in the computer system is determined. The resource scheduler determines the entropy of multiple LC type applications by obtaining the amount of interference that each LC type application can tolerate and the amount of interference that each LC type application actually receives, and according to the amount of interference that each LC type application can tolerate and the amount of interference that each LC type application actually receives. The entropy of at least one BE type application is determined by obtaining the first number of instructions per cycle when each BE type application runs alone and the second number of instructions per cycle after each BE type application is disturbed, and according to the first number of instructions per cycle and the second number of instructions per cycle. Then, the current system entropy is determined according to the entropies of the multiple LC type applications and the entropy of at least one BE type application. Since the importance of the service quality of the LC type application is much greater than that of the BE type application, the weight of the entropy of the plurality of LC type applications is greater than the weight of the entropy of at least one BE type application.

In some optional embodiments of the first aspect, in addition to calculating the current system entropy by itself, the resource scheduler may also obtain the current system entropy from the computing device, which is not specifically limited here.

In an embodiment of the present application, for different types of applications included in the computer system, the resource scheduler can determine the current system entropy in different ways to evaluate the performance of the computer system, and can also flexibly apply to different scenarios, thereby improving the flexibility and practicality of the technical solution of the present application.

In some optional embodiments of the first aspect, the resource scheduler may be triggered to perform resource scheduling in other ways. The resource scheduler obtains the first scheduling information or the second scheduling information from the service quality predictor, wherein the first scheduling information indicates to increase the isolation area resources of the target LC type application among the multiple LC type applications, and the second scheduling information indicates to reduce the isolation area resources of the target LC type application. Afterwards, the resource scheduler increases the isolation area resources of the target LC type application according to the first scheduling information; or reduces the isolation area resources of the target LC type application according to the second scheduling information.

In the embodiment of the present application, there are multiple ways to trigger the resource scheduler to perform resource scheduling. In addition to the remaining tolerance trigger, it can also be triggered based on the remaining tolerance and the current system entropy. In addition, it can also be triggered according to the indication information, which can adapt to different needs in actual applications and improve the flexibility and adaptability of the technical solution. In addition, the resource scheduler can directly determine the area for resource scheduling based on the scheduling information, so that even if the resource scheduler cannot obtain the remaining tolerance or the current system entropy, it can still perform resource scheduling, ensuring the smooth operation of the resource scheduler and improving the reliability of the technical solution.

In some optional embodiments of the first aspect, the resource scheduler can calculate or obtain the system entropy after the computer system resources are scheduled from the computing device, and judge the size relationship between the system entropy after the resource scheduling and the current system entropy. If the system entropy after the resource scheduling is less than the current system entropy, it means that the performance of the system is improved, and it can be determined that the resource scheduling is successful.

In the embodiment of the present application, verification is performed to ensure that resource scheduling will not cause worse results, that is, will not aggravate interference between applications, thereby improving the reliability of the technical solution.

The second aspect of the embodiment of the present application provides a resource scheduling method for a cloud computing system, which is applied to a service quality predictor, and the method includes: the service quality predictor obtains multiple network receiving queue lengths corresponding to the target LC type application, and calculates the average of the multiple network receiving queue lengths to obtain the average network receiving queue length. When the average network receiving queue length is greater than the length threshold, a first scheduling information is sent to the resource scheduler, and the first scheduling information indicates to increase the isolation area resources of the target LC type application. When the average network receiving queue length is less than or equal to the length threshold, and the proportion of the network receiving queue length with a value of 0 in the multiple network receiving queue lengths is greater than the proportion threshold, a second scheduling information is sent to the resource scheduler, and the second scheduling information indicates to reduce the isolation area resources of the target LC type application. Among them, the length threshold and the proportion threshold can be determined by the service quality predictor according to the historical operating status of the computing system, and can also be set by the user, which is not specifically limited here.

In the embodiment of the present application, the service quality predictor monitors the buffer corresponding to each LC type application, and can determine whether resource scheduling is required without calculating system entropy or tolerance, etc., thereby simplifying the process and saving computing resources.

The third aspect of the embodiment of the present application provides an application identification method, which is applied to an application distinguisher, and the method includes: the application distinguisher obtains the average values of multiple network total bandwidths of multiple applications in the current stage in the computer system, and determines the interval variation coefficients of multiple network total bandwidths of multiple applications in the current stage based on the average values of the multiple network total bandwidths. In the case where the interval variation coefficients of the multiple network total bandwidths are greater than the interval coefficient threshold, the multiple absolute values of the difference between the multiple end send/receive bandwidth ratios of the multiple applications in the current stage and the multiple start send/receive bandwidth ratios of the next stage are obtained. The application distinguisher also obtains the multiple send/receive bandwidth ratio variation coefficients of the multiple applications in the target time periods before and after the current stage. Afterwards, according to the multiple absolute values of the difference or the multiple send/receive bandwidth ratio variation coefficients, the identification of the LC type application in the multiple applications is determined as the first application identification, and the identification of the BE type application in the multiple applications is determined as the second application identification. The first application identification and the second application identification are sent to the resource scheduler so that the resource scheduler distinguishes between the BE type application and the LC type application.

In an embodiment of the present application, the application distinguisher can identify the application types of multiple applications in a computer system and inform the resource scheduler of the application identifier of each application, providing technical support for the resource scheduler to perform resource scheduling and improving the feasibility of the technical solution.

In some optional embodiments of the third aspect, according to multiple absolute difference values or multiple transmission/reception bandwidth ratio variation coefficients, determining the identifier of the LC type application in multiple applications as the first application identifier and the identifier of the BE type application in multiple applications as the second application identifier includes: determining from multiple applications that the absolute difference value is greater than the difference threshold, and/or the transmission/reception bandwidth ratio variation coefficient is greater than the coefficient threshold, as the BE type application. Marking the identifier of the BE type application as the second application identifier, and marking the identifiers of the applications other than the BE type application in the multiple applications as the first application identifier.

In the embodiment of the present application, there are many cases for determining the application type. In some cases, the requirements for the distinguishing function of the application distinguisher are not very high, and a certain fault tolerance is allowed. In other cases, it is relatively strict, which is conducive to clearly distinguishing LC type applications from BE type applications and improving the reliability of the system.

A fourth aspect of an embodiment of the present application provides a resource scheduling method for a cloud computing system, which is applied to a resource scheduler. The method includes: obtaining first scheduling information or second scheduling information from a service quality predictor, the first scheduling information indicating to increase the isolation area resources of a target LC type application among multiple LC type applications, and the second scheduling information indicating to reduce the isolation area resources of the target LC type application; increasing the isolation area resources of the target LC type application according to the first scheduling information; or reducing the isolation area resources of the target LC type application according to the second scheduling information.

In an embodiment of the present application, the resource scheduler can directly determine the resource scheduling area based on the scheduling information, so that even if the resource scheduler cannot obtain the remaining tolerance or the current system entropy, it can still perform resource scheduling, thereby ensuring the smooth operation of the resource scheduler and improving the reliability of the technical solution.

In a fifth aspect, an embodiment of the present application provides a resource scheduling system, the resource scheduling system comprising a resource scheduler, the resource scheduler being used to: obtain the residual interference tolerance of each LC type application in at least one delay-sensitive LC type application included in the computer system. From multiple LC type applications, obtain a first LC type application with the smallest residual interference tolerance. If the residual interference tolerance of the first LC type application is less than the lower limit of the tolerance, increase the first isolation area resources of the first LC type application. If the residual interference tolerance of the first LC type application is greater than the upper limit of the tolerance, transfer the second isolation area resources of the second LC type application in the multiple LC type applications to the resource sharing area.

In some optional embodiments of the fifth aspect, the resource scheduler is further used to: obtain the current system entropy of the computer system, the current system entropy is used to currently indicate the degree of interference between applications in the computer system. If the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is less than the lower limit of the tolerance, then increase the first isolation area resources. If the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is greater than the upper limit of the tolerance, then transfer the second isolation area resources of the second LC type application to the resource sharing area.

In some optional embodiments of the fifth aspect, the resource scheduler is specifically configured to: if there is a third LC type application in the computer system whose remaining interference tolerance is greater than the tolerance upper limit and has a third isolated area resource that can be stripped, transfer resources in the third isolated area resources to the first isolated area corresponding to the first LC type application to increase the resources of the first isolated area. If there is no third LC type application in the computer system, transfer resources in the resource sharing area to the first isolated area to increase the resources of the first isolated area.

In some optional embodiments of the fifth aspect, the resource scheduler is specifically used to: obtain the amount of interference that each LC type application can tolerate and the amount of interference actually received by each LC type application. According to the amount of interference that each LC type application can tolerate and the amount of interference actually received by each LC type application, determine the current system entropy.

In some optional embodiments of the fifth aspect, the resource scheduler is further used to: obtain a first application identifier and a second application identifier from an application distinguisher or from a user, the first application identifier being used to indicate an LC type application, and the second application identifier being used to indicate a BE type application. Determine the LC type application in the computer system based on the first application identifier. Determine the BE type application in the computer system based on the second application identifier.

The resource scheduler is specifically used to: obtain the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives. Determine the entropy of multiple LC-type applications based on the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives. Obtain a first number of instructions per cycle when each BE-type application of at least one BE-type application runs alone and a second number of instructions per cycle after each BE-type application is interfered with. Determine the entropy of at least one BE-type application based on the first number of instructions per cycle and the second number of instructions per cycle. Determine the current system entropy based on the entropies of multiple LC-type applications and the entropy of at least one BE-type application.

In some optional embodiments of the fifth aspect, the resource scheduler is further used to: obtain first scheduling information or second scheduling information from the service quality predictor, the first scheduling information indicating to increase the isolation area resources of the target LC type application among the multiple LC type applications, and the second scheduling information indicating to reduce the isolation area resources of the target LC type application. According to the first scheduling information, the isolation area resources of the target LC type application are increased. Alternatively, according to the second scheduling information, the isolation area resources of the target LC type application are reduced.

In some optional embodiments of the fifth aspect, the resource scheduler is further used to: obtain system entropy after resource scheduling of the computer system. If the system entropy after resource scheduling is less than the current system entropy, it is determined that the resource scheduling is successful.

In some optional embodiments of the fifth aspect, the resource scheduling system further includes a service quality predictor, which is used to: obtain multiple network receive queue lengths corresponding to the target LC type application. Calculate the average of the multiple network receive queue lengths to obtain the average network receive queue length. If the average network receive queue length is greater than the length threshold, send first scheduling information to the resource scheduler, and the first scheduling information indicates to increase the isolation area resources of the target LC type application. If the average network receive queue length is less than or equal to the length threshold, and the proportion of the network receive queue length with a value of 0 in the multiple network receive queue lengths is greater than the proportion threshold, send second scheduling information to the resource scheduler, and the second scheduling information indicates to reduce the isolation area resources of the target LC type application.

In some optional embodiments of the fifth aspect, the resource scheduling system further includes an application distinguisher, which is used to: obtain multiple average network total bandwidths of multiple applications in the computer system at the current stage. According to the multiple average network total bandwidths, determine the interval variation coefficients of the multiple network total bandwidths of the multiple applications at the current stage. If the multiple interval variation coefficients of the multiple network total bandwidths are greater than the interval coefficient threshold, obtain multiple absolute values of the difference between the multiple end send/receive bandwidth ratios of the multiple applications in the current stage and the multiple start send/receive bandwidth ratios of the next stage. Obtain multiple send/receive bandwidth ratio variation coefficients of the multiple applications in the target time periods before and after the current stage. According to the multiple absolute values of the difference or the multiple send/receive bandwidth ratio variation coefficients, determine that the identifier of the LC type application in the multiple applications is the first application identifier, and the identifier of the BE type application in the multiple applications is the second application identifier. Send the first application identifier and the second application identifier to the resource scheduler so that the resource scheduler distinguishes between the BE type application and the LC type application.

In some optional embodiments of the fifth aspect, the application distinguisher is specifically used to: determine, from multiple applications, an application whose absolute value of difference is greater than a difference threshold, and/or whose coefficient of variation of the transmission/reception bandwidth ratio is greater than a coefficient threshold, as a BE type application. Mark the identifier of the BE type application as the second application identifier. Mark the identifier of an application other than the BE type application among the multiple applications as the first application identifier.

In some optional embodiments of the fifth aspect, the resource scheduling system also includes a computing device, which is used to send to the resource scheduler at least one of the current system entropy, or the amount of interference that each LC-type application can tolerate, or the amount of interference actually received by each LC-type application, or the first number of instructions per cycle when each BE-type application runs alone, or the second number of instructions per cycle after each BE-type application is interfered.

In some optional embodiments of the fifth aspect, the computing device is further configured to obtain a plurality of network receive queue lengths corresponding to the target LC type application, calculate an average of the plurality of network receive queue lengths to obtain an average network receive queue length, and send the average network receive queue length to the service quality predictor.

In some optional embodiments of the fifth aspect, the computing device is also used to send to the application distinguisher an average value of multiple network total bandwidths of multiple applications in the computer system at the current stage, or multiple network total bandwidth interval variation coefficients of multiple applications in the current stage, or multiple absolute values of differences between multiple end send/receive bandwidth ratios of multiple applications in the current stage and multiple starting send/receive bandwidth ratios of the next stage, or at least one of multiple send/receive bandwidth ratio variation coefficients of multiple applications in the target time periods before and after the current stage.

The resource scheduling system is used to implement the method shown in any one of the first to fourth aspects mentioned above. The beneficial effects of the fifth aspect are similar to those of the first to fourth aspects and will not be repeated here.

A sixth aspect of an embodiment of the present application provides a resource scheduler, which is applied to a cloud computing system. The resource scheduler includes:

The acquisition unit is configured to acquire the residual interference tolerance of each LC type application in at least one delay-sensitive LC type application included in the computer system, and to acquire a first LC type application with the smallest residual interference tolerance from the plurality of LC type applications.

A processing unit for:

If the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, the first isolation area resources of the first LC type application are increased.

If the remaining interference tolerance of the first LC-type application is greater than the tolerance upper limit, the second isolation area resources of the second LC-type application among the multiple LC-type applications are transferred to the resource sharing area.

The resource scheduler is used to implement the method of the first aspect. The beneficial effects of the sixth aspect are similar to those of the first aspect and will not be repeated here.

A seventh aspect of an embodiment of the present application provides a service quality predictor, which is applied to a cloud computing system. The service quality predictor includes:

The acquisition unit is used to acquire the lengths of multiple network receiving queues corresponding to the target LC type application.

The processing unit is used to calculate the average of the lengths of multiple network receiving queues to obtain the average network receiving queue length.

A sending unit, for:

If the average network receiving queue length is greater than the length threshold, first scheduling information is sent to the resource scheduler, where the first scheduling information indicates to increase the isolation area resources of the target LC type application.

If the average network receive queue length is less than or equal to the length threshold, and the proportion of network receive queue lengths with a value of 0 among multiple network receive queue lengths is greater than the proportion threshold, then a second scheduling information is sent to the resource scheduler, and the second scheduling information indicates to reduce the isolation area resources of the target LC type application.

The service quality predictor is used to implement the method of the second aspect. The beneficial effects of the seventh aspect are similar to those of the second aspect and will not be repeated here.

An eighth aspect of the embodiments of the present application provides an application distinguisher, including:

The acquisition unit is used to acquire average values of multiple network total bandwidths of multiple applications in the computer system at the current stage.

The processing unit is used to determine multiple network total bandwidth interval variation coefficients of multiple applications at a current stage according to multiple network total bandwidth average values.

The acquisition unit is further used to acquire the absolute values of the differences between the multiple end sending/receiving bandwidth ratios of the multiple applications in the current stage and the multiple starting sending/receiving bandwidth ratios in the next stage if the multiple network total bandwidth interval variation coefficients are greater than the interval coefficient threshold.

The acquisition unit is further used to acquire multiple sending/receiving bandwidth ratio variation coefficients of multiple applications in the target time periods before and after the current stage.

The processing unit is further used to determine, based on multiple difference absolute values or multiple transmission/reception bandwidth ratio variation coefficients, an identifier of an LC type application among multiple applications as a first application identifier and an identifier of a BE type application among multiple applications as a second application identifier.

The sending unit is used to send the first application identifier and the second application identifier to the resource scheduler, so that the resource scheduler can distinguish between the BE type application and the LC type application.

The application distinguisher is used to implement the method of the third aspect. The beneficial effects of the eighth aspect are similar to those of the third aspect and will not be repeated here.

A ninth aspect of an embodiment of the present application provides a resource scheduler, which is applied to a cloud computing system. The resource scheduler includes:

The acquisition unit is used to acquire first scheduling information or second scheduling information from the service quality predictor, the first scheduling information indicates to increase the isolation area resources of the target LC type application among multiple LC type applications, and the second scheduling information indicates to reduce the isolation area resources of the target LC type application.

The processing unit is further configured to increase the isolation area resources of the target LC type application according to the first scheduling information; or reduce the isolation area resources of the target LC type application according to the second scheduling information.

The resource scheduler is used to implement the method of the fourth aspect. The beneficial effects of the ninth aspect are similar to those of the fourth aspect and will not be repeated here.

The tenth aspect of the embodiment of the present application provides a computer device, including a processor, a memory and a communication interface, the processor, the memory and the communication interface are connected, and the processor is used to execute the method of any one of the first to fourth aspects. The beneficial effects shown in this aspect are similar to those of the first to fourth aspects, and will not be repeated here.

An eleventh aspect of the embodiments of the present application provides a computer-readable storage medium, in which a program is stored. When a computer executes the program, the method of any one of the first to fourth aspects is executed.

A twelfth aspect of the embodiments of the present application provides a computer program product, characterized in that when the computer program product is executed on a computer, the computer executes the method of any one of the first to fourth aspects.

The beneficial effects of the eleventh and twelfth aspects are similar to those of the first to fourth aspects and will not be repeated here.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1a is a schematic diagram of an isolated resource area and a shared resource area;

FIG1 b is another schematic diagram of an isolated resource area and a shared resource area;

FIG2 is a schematic diagram of a system architecture of a high-concurrency cloud computing system;

FIG3 is a schematic diagram of a flow chart of a resource scheduling method provided in an embodiment of the present application;

FIG4 is another schematic diagram of a flow chart of a resource scheduling method provided in an embodiment of the present application;

FIG5 is another schematic diagram of a flow chart of a resource scheduling method provided in an embodiment of the present application;

FIG6 is another schematic diagram of a flow chart of a resource scheduling method provided in an embodiment of the present application;

FIG7 is another schematic diagram of a flow chart of a resource scheduling method provided in an embodiment of the present application;

FIG8 is a schematic diagram of a code provided in an embodiment of the present application;

FIG9 is another schematic diagram of a flow chart of a resource scheduling method provided in an embodiment of the present application;

FIG10a is a schematic diagram of the transmission bandwidth ratio of the Redis application;

FIG10 b is a schematic diagram of the transmission bandwidth ratio of the Terasort application based on the Spark framework;

FIG11 is a flow chart of an application identification method provided in an embodiment of the present application;

FIG12 is a flow chart of an application identification method provided in an embodiment of the present application;

FIG13a is a schematic diagram of a structure of a resource scheduling system provided in an embodiment of the present application;

FIG13b is another schematic diagram of the structure of the resource scheduling system provided in an embodiment of the present application;

FIG14 is a schematic diagram of a structure of a resource scheduler provided in an embodiment of the present application;

FIG15 is a schematic diagram of a structure of a service quality predictor provided in an embodiment of the present application;

FIG16 is a schematic diagram of a structure of an application distinguisher provided in an embodiment of the present application;

FIG. 17 is a schematic diagram of the structure of a computer device provided in an embodiment of the present application.

DETAILED DESCRIPTION

The embodiments of the present application provide a resource scheduling method, an application identification method and related equipment for a cloud computing system. In the resource scheduling method, different resource scheduling modes are determined according to different values of the residual interference tolerance of each LC-type application in the computer system, so as to increase the isolation area resources of the LC-type application, or reduce the isolation area resources of the LC-type application to increase the shared area resources of the computer system, thereby realizing flexible allocation of isolation area resources and shared area resources to adapt to the actual needs of the LC-type application, avoid resource waste, and improve system performance.

The embodiments of the present application are described below in conjunction with the accompanying drawings. It is known to those skilled in the art that with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged in appropriate circumstances, which is only to describe the distinction mode adopted by the objects of the same attribute in the embodiments of the present application when describing. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment containing a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment. In addition, "at least one" refers to one or more, and "multiple" refers to two or more. "And/or", describes the association relationship of associated objects, indicating that three relationships can exist, for example, A and/or B, can represent: A exists alone, A and B exist simultaneously, and B exists alone, wherein A, B can be singular or plural. The character "/" generally represents that the associated objects before and after are a kind of "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or plural.

First, the proper nouns and related concepts that may be involved in the embodiments of the present application are explained.

1. Isolation area resources and shared area resources.

The resources that can be scheduled in each application are divided into two areas: the isolation area and the shared area. The resources in the isolation area are called isolation area resources, and the resources in the shared area are called shared area resources. The isolation area can also be called the isolation resource area, which refers to the resource area that can be used by the application to which the isolation area belongs. In other words, only the application to which the isolation area belongs can use the resources in an isolation area, and other applications cannot use them. The shared area can also be called the shared resource area, which is a resource area that can be used by all applications. In other words, the shared area resources are resources that can be used by all applications. In a system, there is only one shared area.

For a clearer explanation, please refer to FIG. 1a and FIG. 1b , which are schematic diagrams of an isolated resource area and a shared resource area.

As shown in Figure 1a, before resource scheduling, there is only a shared resource area in the system, but no isolated resource area. That is to say, if the service quality of application 1 and application 2 is reduced, that is, application 1 and application 2 are interfered by other applications, resource scheduling is required to allocate part of the resources in the shared resource area to application 1 and application 2. Then, after resource scheduling, application 1 and application 2 have exclusive isolated area resources. This reduces the interference to application 1 and application 2.

As shown in FIG. 1 b , in some optional embodiments, each application may have an exclusive isolated resource area and also use resources in a shared resource area.

It should be noted that, in actual applications, since the service quality of LC applications is much more important than that of BE applications, in the embodiment of the present application, LC applications are set as critical applications and BE applications are set as non-critical applications. Therefore, the isolation area resources of BE applications can be set to always be 0.

2. Latency critical (LC) applications and best-efforts (BE) applications.

The service quality of LC applications is usually measured by tail latency (TL). When the service rate of LC applications is slower than the mandatory request arrival rate, a large number of requests will be queued for processing, resulting in a sharp increase in request queuing time, which further leads to a sharp increase in tail latency. In other words, interference from other applications will have a very destructive impact on the tail latency of LC applications.

As for BE type applications, in actual applications, the main concern is its execution time, and the performance degradation caused by interference to BE type applications will not bring fatal impact. Among them, BE type applications can also be called batch processing applications.

3. Highly concurrent cloud computing system.

In actual applications, in order to ensure the service quality of applications and improve the resource utilization of machines, LC-type applications and BE-type applications are usually deployed in a mixed manner on the same physical machine, equipped with multiple virtual machines, to form a high-concurrency cloud computing system. Among them, high concurrency refers to the design that ensures that the system can process a large number of requests in parallel at the same time, which is a factor that must be considered in the design of Internet distributed system architecture. Cloud computing is a type of distributed computing, which refers to dividing huge data computing processing programs into multiple small programs through the network cloud, and then processing them through a system composed of multiple servers to obtain results and return them to users.

The high-concurrency cloud computing system is described below in conjunction with Figure 2. Please refer to Figure 2, which is a schematic diagram of a system architecture of the high-concurrency cloud computing system.

As shown in FIG2 , in a high-concurrency cloud computing system, a physical machine is equipped with multiple virtual machines, and each virtual machine runs one or more applications. The applications running on the virtual machines include LC-type applications and BE-type applications. The multiple applications running on the same virtual machine can be applications of the same type or different types, which are not specifically limited here. For example, in the embodiment shown in FIG2 , application 2 and application 3 running on virtual machine 2 can both be LC-type applications or BE-type applications, or one can be an LC-type application and the other can be a BE-type application. Multiple virtual machines can run their respective applications at the same time, so that the system can process a large number of requests in parallel at the same time, ensuring high concurrency functions.

Next, please refer to FIG. 3, which is a flow chart of a resource scheduling method provided in an embodiment of the present application, including the following steps:

301. Obtain the residual interference tolerance of each LC type application in at least one delay-sensitive LC type application included in the computer system.

Assume that the computer system includes N LC applications, where N is a positive integer greater than 2. For each LC application, the resource scheduler can obtain the following variables: 1) the ideal tail delay TL _i0 of user i, that is, the tail delay when a single LC application is running; 2) the tail delay TL _i1 of user i after interference, that is, the tail delay when multiple applications are running together; 3) the maximum tolerable tail delay M _i of user i, whose value is determined by the user. Among these three variables, TL _i0 <M _i , TL _i0 <TL _i1 .

According to the above variables, based on formula 1, the amount of interference that user i can tolerate (anti-interference capability) is defined as:

Formula 1:

Wherein, _Ai represents the amount of interference that user i can tolerate. Since TL _i0 ＜M _i , the value range of _Ai is between (0, 1). The smaller _{Mi is} , the closer _Ai is to 0, indicating that the anti-interference ability of user i is weaker; the larger _Mi is, the closer _Ai is to 1, indicating that the anti-interference ability of user i is stronger.

According to the above variables, based on Formula 2, the amount of interference that user i actually suffered is defined as:

Formula 2:

Wherein, _Ri represents the actual interference amount to which user i is subjected. Since TL _i0 < TL _i1 , the value range of _Ri is between (0, 1). The smaller TL _i1 is, the closer _Ri is to 0, indicating that the actual interference to user i is smaller; the larger TL _i1 is, the closer _Ri is to 1, indicating that the actual interference to user i is greater.

According to the above variables, the residual interference tolerance S _i of LC type application i is defined based on formula 3:

Formula 3:

Combining Formula 1, Formula 2 and Formula 3, the resource scheduler can determine the residual interference tolerance of each LC type application in the computer system. The residual interference tolerance reflects the residual anti-interference ability of the application. The larger the residual interference tolerance, the less interference the application is subjected to and the stronger the residual anti-interference ability is.

302. Obtain a first LC type application with the smallest residual interference tolerance from multiple LC type applications.

After the residual interference tolerance of each LC type application is acquired, the residual interference tolerance of each LC type application may be compared to determine a first LC type application having the smallest residual interference tolerance.

303. If the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, increase the first isolation area resources of the first LC type application.

The resource scheduler will also compare the remaining interference tolerance of the first LC type application with the tolerance lower limit to determine whether the interference level of the first LC type application exceeds the tolerance threshold of the application. If the remaining interference tolerance lower limit of the first LC type application is less than the tolerance lower limit, it means that the first LC type application is seriously interfered and will violate the quality of service (QoS) target. Then it is necessary to increase the first isolation area resources of the first LC type application to reduce the interference of other applications to the first LC type application.

In the embodiments of the present application, there are multiple ways to increase the resources of the first isolation area, and the possible situations are described below respectively.

It should be noted that, as explained above, the shared area resources are resources used by all applications in the computer system. Therefore, if resource scheduling is required, if the resources in the shared area are directly transferred to the isolated area of the first LC type application, it will have a greater impact on the performance of the system. Therefore, it is preferred to transfer resources from the isolated areas of other LC type applications.

1) Transfer resources from the isolation area of other applications to the isolation area of the first LC type application.

In this case, the resource scheduler will select the resource-rich LC type application for resource transfer. If the resource scheduler determines that there is a third LC type application in the computer system whose remaining interference tolerance is greater than the tolerance upper limit, and the third LC type application has a third isolation area resource that can be stripped, it means that while ensuring the normal operation of the third LC type application, the isolation area resources of the third LC type application still have excess resources, and part or all of the third isolation area resources can be transferred to the first isolation area to increase the resources of the first isolation area.

2) Transfer resources from the shared area to the isolated area of the first LC type application.

If the resource scheduler determines that the third LC type application does not exist in the computer system, the resource scheduler will transfer the resources of the resource sharing area to the first isolated area to increase the resources of the first isolated area.

Specifically, there are the following situations:

One situation is that the resource scheduler determines that there are LC type applications in the computer system whose remaining interference tolerance is greater than the tolerance upper limit, but the isolation area resources of these LC type applications are not separable.

Another situation is that in a computer system, the remaining interference tolerance of all LC-type applications is less than or equal to the upper tolerance limit, that is, the isolation area resources of all LC-type applications can only ensure that the application is in a normal operating state. If its own isolation area resources are transferred to other LC-type applications, its own operation will be affected, exacerbating the interference between applications.

Exemplarily, the resources transferred in the embodiments of the present application include divisible hardware resources, including the number of central processing units (CPU) cores, the number of last-level shared cache paths, etc. In addition, it can also be other hardware resources, such as memory bandwidth, disk bandwidth, etc., which are not specifically limited here.

In the embodiment of the present application, when increasing the first isolation area resources of the first LC type application, priority is given to transferring resources from the isolation area resources of other LC type applications, so as to minimize the adverse effects on system performance and improve the practicality of the technical solution. At the same time, for different situations, there are different ways to increase the first isolation area resources, which can adapt to different scenarios and improve the flexibility and adaptability of the technical solution of the present application.

304. If the remaining interference tolerance of the first LC type application is greater than the tolerance upper limit, the second isolation area resources of the second LC type application among the multiple LC type applications are transferred to the resource sharing area.

Since the remaining interference tolerance of the first LC type application is the smallest among the remaining tolerances of the multiple LC type applications, if the remaining interference tolerance of the first LC type application is greater than or equal to the tolerance upper limit, this means that each LC type application included in the system has abundant resources and is not interfered with or is interfered with to a lesser extent. Then, an LC type application with abundant resources can be selected from the multiple LC type applications for resource stripping, that is, the resources of the second isolation area of the second LC type application are transferred to the resource sharing area.

The second LC type application refers to an application whose residual interference tolerance is greater than that of the first LC type application. In some optional embodiments, the LC type application with the largest residual interference tolerance (i.e., the most abundant resources) among multiple LC type applications can be selected as the second LC type application for resource stripping. Selecting the application with the most abundant resources for resource stripping can ensure that the utilization rate of resources is improved without affecting the normal operation of the application.

In the embodiment of the present application, different resource scheduling modes are determined according to different values of the residual interference tolerance of each LC type application in the computer system, so as to increase the isolated area resources of the LC type application, or reduce the isolated area resources of the LC type application to increase the shared area resources of the computer system, thereby realizing the flexible allocation of isolated area resources and shared area resources and improving the utilization rate of resources. At the same time, by scheduling isolated area resources and shared area resources, the interference between applications is reduced and the system performance is also improved.

In some optional embodiments, since the resources of the first isolation area in step 203 are increased, the first isolation area of the first LC type application is called the beneficiary area, and the area to which the resources are transferred to the first isolation area is called the victim area. Similarly, the resources of the second isolation area of the second LC type application in step 204 can be transferred to the shared area, and therefore, the second isolation area corresponding to the second LC type application is called the victim area, and the shared area is called the beneficiary area.

The embodiment shown in Figure 3 is further described below from the perspective of the victim area and the benefited area. Please refer to Figures 4 to 6, which are all flowchart diagrams of the resource scheduling method provided in the embodiments of the present application.

Please refer to FIG4 . The embodiment shown in FIG4 is a process of selecting a benefited area, which includes the following steps:

401. Determine a first LC type application having the smallest residual interference tolerance from a plurality of LC type applications.

Step 401 is similar to step 302 in the embodiment shown in FIG. 3 , which has been described above and will not be repeated here.

402. Whether the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, if so, execute step 403, if not, execute step 404.

The resource scheduler compares the remaining interference tolerance of the first LC type application with the tolerance lower limit to determine whether the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, so as to determine whether the isolation area of the first LC type application is a benefit area according to the comparison result. The tolerance lower limit can be set by the user according to the system requirements, or by the resource scheduler or other devices, which is not limited here.

403. Return to the first isolation area corresponding to the first LC type application.

If the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, it means that the first LC type application is seriously interfered and the isolation area resources of the first LC type application need to be increased. Therefore, the first isolation area corresponding to the first LC type application can be determined as the benefited area.

404. Return to shared area.

If the remaining interference tolerance of the first LC type application is not less than the tolerance lower limit, it means that the first LC type application with the smallest remaining tolerance is less interfered or not interfered, and can operate normally, and there is no need to increase the isolation area resources of any LC type application. In this case, it is possible to choose to strip resources from the isolation area of the resource-rich LC type application to the shared area, so the shared area can be determined as the benefited area.

The resource scheduler may also determine the victim area. Please refer to FIG. 5 . The embodiment shown in FIG. 5 is a process of selecting the victim area, including the following steps:

501. Is there a third LC type application among the multiple LC type applications whose remaining interference tolerance is greater than the tolerance upper limit and has a strippable isolation area resource? If so, execute step 502; if not, execute step 503.

The resource scheduler determines whether there is a third LC type application among the multiple LC type applications that is less interfered or not interfered and has separable isolation area resources. The interference degree is determined by the residual interference tolerance, and the greater the residual interference tolerance, the smaller the interference degree.

The upper limit of tolerance may be set by the user according to system requirements, or may be set by a resource scheduler or other device, which is not specifically limited here.

In some optional embodiments, if there are multiple third LC type applications, the resource scheduler may preferentially select an application with the largest remaining interference tolerance as the application to transfer resources, which is not specifically limited here.

502. Return to the third isolation area corresponding to the third LC type application.

If there is a third LC type application, the resource scheduling may determine a third isolation area corresponding to the third LC type application as a victim area.

503. Return to shared area.

If there is no third LC type application, it means that the remaining tolerance of all LC type applications is below the upper tolerance limit. The resources of each LC type application can only ensure its own normal operation and cannot transfer resources to other applications. Therefore, the shared area is selected as the victim area.

In the embodiment of the present application, in addition to clarifying the victim area and the beneficiary area, the resource type transferred from the victim area to the beneficiary area can also be determined, and it can be determined whether the transfer can be successful. That is, in the embodiment shown in FIG6, multiple functions are called to transfer resources. Among them, the embodiment shown in FIG6 can be called to call the resource AdjustResource function to adjust resource allocation. Please refer to FIG6, which includes the following steps:

601. Determine the victim region through findVictimRegion function.

The embodiment shown in FIG5 is a specific process of determining the victim area, as shown in FIG5 for details, which will not be described again here.

602. Determine the beneficiary area through the findBeneficiaryRegion function.

The embodiment shown in FIG4 is a specific process of determining the benefited area, as shown in FIG4 for details, which will not be described again here.

It should be noted that there is no necessary order between step 601 and step 602. In actual application, step 601 may be executed first, or step 602 may be executed first, or step 601 and step 602 may be executed simultaneously. The specific order is not limited here.

603. Determine the resources to be adjusted through the findVictimRegion function.

In the findVictimRegion function, a finite state machine is maintained for each application. Each state of the state machine represents the current resource type, and the resource type to be adjusted in the victim area is determined based on the state of the finite state machine. When condition 1 is met: the current resource type can no longer be stripped; or condition 2: the adjustment of the current resource type will cause the system entropy of the computer system to increase (that is, increase the interference between applications), the state of the state machine will transfer. The state machine transfer of the state machine means that the type of resource to be stripped changes. For example, assuming that there are 3 resource types in the victim area, and the current resource type is resource 1, when any one of the above conditions is met, the state of the state machine transfers, and when the isolation area resources are stripped, the stripped resource is no longer resource 1, but resource 2.

604. Move the resources to be adjusted by a unit from the affected area to the benefited area.

After determining the victim area and the beneficiary area, the resource scheduler will transfer the resources to be adjusted by a unit from the victim area to the beneficiary area to realize resource scheduling.

It should be noted that in actual applications, the victim area and the beneficiary area may be the same area, in which case no resource scheduling will be performed. There are two reasons for this situation: one is that the current resource allocation meets the demand, that is, the remaining interference tolerance of all LC-type applications is between the lower tolerance limit and the upper tolerance limit; the other is that all LC-type applications have no isolation area that can be stripped.

605. Are the resources adjusted successfully? If so, execute step 606; if not, execute step 607.

After resource scheduling, the resource scheduler will determine the benefits of the adjustment based on the change in system entropy, and choose to keep the change or roll it back. System entropy indicates the degree of interference between applications in a computer system.

Specifically, the resource scheduler can obtain the system entropy before and after scheduling. If the system entropy increases after resource scheduling, it is determined that the resource scheduling is unsuccessful. If the system entropy decreases after resource scheduling, it is determined that the resource scheduling is successful.

606. Returns "True".

When it is determined that the resources are successfully adjusted, the resource scheduler may also return a "True" result, indicating that the resource scheduling is successful.

607. Returns "False".

In the case of determining that the resource adjustment fails, the resource scheduler may also return a "False" result, indicating that the resource scheduling is unsuccessful, canceling the last adjustment and prohibiting the stripping of the current victim area resources in the next period of time.

In some optional embodiments, in addition to comparing the remaining tolerance of the first LC type application with the tolerance lower limit as shown above, the condition for determining whether to perform resource scheduling may also include comparing the current system entropy of the computer system with the system entropy threshold, and combining the two comparison results to determine whether to perform resource scheduling. This situation is described below.

Simply put, the resource scheduler can obtain the current system entropy of the computer system, and the current system entropy is used to indicate the degree of interference between applications in the computer system. The larger the current system entropy, the greater the degree of interference between applications in the computer system. Therefore, when the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is less than the lower limit of the tolerance, the resource scheduler will increase the resources of the first isolation area. When the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is greater than the upper limit of the tolerance, the resource scheduler will transfer the second isolation area resources of the second LC type application to the resource sharing area. The specific way in which the resource scheduler transfers resources has been explained in the embodiment shown in Figure 3 and will not be repeated here.

The system entropy threshold may be set by a user, or by a resource scheduler or other device in combination with a historical operating state of the computer system, and is not specifically limited here.

In the embodiment of the present application, the remaining tolerance and system entropy are combined to determine whether to perform resource scheduling, and the conditions for resource scheduling are more strictly limited, so as to avoid performing resource scheduling in unnecessary situations and save computing resources. Among them, the unnecessary situation includes that although the remaining tolerance of the first LC type application is lower than the tolerance lower limit, the current system entropy is less than the system entropy threshold, that is, the computer system as a whole is still in normal operation, and not performing resource scheduling will not cause a big problem.

Optionally, there are many ways for the resource scheduler to obtain the current system entropy. The resource scheduler can calculate the current system entropy by itself, or receive the current system entropy from other devices (such as computing devices). The specifics are not limited here. The following is an example of the resource scheduler calculating by itself for explanation.

Since there are many possible applications included in the computer system, the resource scheduler can determine the current system entropy in different ways. Different situations are described below.

1) The computer system includes only LC type applications.

In this case, the resource scheduler obtains the amount of interference that each LC type application can tolerate and the amount of interference that each LC type application actually experiences, and then determines the current system entropy based on the amount of interference that each LC type application can tolerate and the amount of interference that each LC type application actually experiences.

Specifically, take the assumption in step 301 that the computer system includes N LC applications as an example. The resource scheduler obtains the amount of interference _Ai that each _Ri application can tolerate according to formula 1, and obtains the amount of interference _Ri actually received by each LC application according to formula 2. Formulas 1 and 2 are as follows:

Formula 1: Formula 2:

Wherein, TL _i0 represents the ideal tail delay of user i, TL _i1 represents the tail delay of user i after interference, and _Mi represents the maximum tolerable tail delay of user i. Among these three variables, TL _i0 <M _i , TL _i0 <TL _i1 .

According to the above variables, based on Formula 4, the resource scheduler can determine the interference Qi that is beyond the tolerance of user _i :

Formula 4:

Based on Formula 5, the resource scheduler can determine the system entropy E _LC in a computer system including only N LC-type applications:

Formula 5:

Here, E _LC represents the amount of interference that LC-type applications cannot tolerate, that is, the system entropy of a computer system with only LC-type applications.

2) Computer systems include LC type applications and BE type applications.

In this case, the resource scheduler obtains a first application identifier and a second application identifier from an application distinguisher or a user, wherein the first application identifier is used to indicate an LC type application, and the second application identifier is used to indicate a BE type application. Then, the LC type application in the computer system is determined according to the first application identifier; and the BE type application in the computer system is determined according to the second application identifier. Then, the system entropy of the LC type application and the system entropy of the BE type application are determined respectively, thereby determining the current system entropy of the computer system.

The process of determining the system entropy of the LC type application has been described above and will not be repeated here. The following describes how to determine the system entropy of the BE type application.

The resource scheduler obtains a first number of instructions per cycle (IPC) of each BE type application when it runs alone and a second number of instructions per cycle of each BE type application after being disturbed, and then determines the system entropy of the BE type application based on the first IPC and the second IPC.

Exemplarily, assuming that the computer system includes M BE type applications, where M is a positive integer, the resource scheduler can determine the system entropy E _BE of the BE type applications according to Formula 6:

Formula 6:

Among them, E _BE represents the amount of interference that BE type applications cannot tolerate; IPC _optional (i) represents the IPC when application i runs alone, that is, the first IPC; IPC _real (i) represents the IPC after application i is interfered, that is, the second IPC.

After obtaining the system entropy of the LC type application and the system entropy of the BE type application, the current system entropy E _S is determined based on Formula 7:

Formula 7: E _S = w × E _LC + (1-w) × E _BE

The value range of w is [0, 1]. Since the service quality of LC type applications is much more important than that of BE type applications, it is also hoped that the lowest E _S can be achieved while reducing E _LC and E _BE . Therefore, the value range of w is set to [0.5, 1], so that E _LC is reduced first, that is, the priority of reducing E _LC is greater than the priority of reducing E _BE .

In an embodiment of the present application, for different types of applications included in the computer system, the resource scheduler can determine the current system entropy in different ways to evaluate the performance of the computer system, and can also flexibly apply to different scenarios, thereby improving the flexibility and practicality of the technical solution of the present application.

In the embodiment of the present application, the resource scheduler will also verify the result of resource scheduling. The resource scheduler can obtain the system entropy after the resource scheduling of the computer system; if the system entropy after the resource scheduling is less than the current system entropy, the resource scheduling is determined to be successful. If the system entropy after the resource scheduling is not less than the current system entropy, the last resource adjustment will be rolled back.

Optionally, the resource scheduler may calculate the system entropy after resource scheduling based on the above formula, and may also receive the system entropy after resource scheduling from other devices (eg, computing devices), which is not specifically limited here.

In the embodiment of the present application, verification is performed to ensure that resource scheduling will not cause worse results, that is, will not aggravate interference between applications, thereby improving the reliability of the technical solution.

In general, whether the triggering of resource scheduling is based on the remaining tolerance or based on the remaining tolerance and the current system entropy, the whole process of resource scheduling can be shown in FIG7, where FIG7 takes the resource scheduler calculating the current system entropy by itself as an example. Please refer to FIG7, which is a flow chart of a resource scheduling method provided in an embodiment of the present application, including the following steps:

701. Monitor the tail delay of LC type applications and the IPC of BE type applications, and calculate the anti-interference ability of LC type applications.

The resource scheduler monitors the tail delay of LC type applications and the IPC of BE type applications to facilitate the calculation of the current system entropy. The anti-interference ability of LC type applications can be reflected by the residual interference tolerance. The specific calculation method is in the above formula and description, which will not be repeated here.

702. Calculate the current system entropy.

The method for calculating the current system entropy by the resource scheduler has been described above and will not be repeated here.

703. Determine the LC type application with the minimum residual tolerance.

Step 703 is similar to step 302 of the embodiment shown in FIG. 3 , and will not be described again here.

704. Perform resource scheduling.

The resource scheduler performs resource scheduling based on the embodiments shown in FIG. 3 to FIG. 6 , and transfers resources from the victim area to the benefited area.

705. Determine whether the resources are successfully adjusted and the system entropy increases. If so, execute step 706; if not, execute step 707.

After the resources are successfully adjusted, it is necessary to determine whether the result of the resource adjustment will lead to an increase in system entropy. According to the judgment result, corresponding operations are performed to ensure that the interference between applications in the system will not increase.

706. Roll back the last resource adjustment.

If the system entropy increases, it means that the interference between applications has increased. It is necessary to roll back the last resource adjustment and prohibit resource stripping from the victim in the next period of time.

707. Call AdjustResource function to adjust resource allocation.

If the system entropy does not increase, the resource adjustment behavior can be retained, that is, the operation of calling the AdjustResource function to adjust resource allocation can be retained.

For example, assuming that the lower tolerance limit is 0.05 and the upper tolerance limit is 0.1, please refer to Figure 8, which is a schematic diagram of a code provided by an embodiment of the present application. The resource scheduler can implement resource scheduling based on the code shown in Figure 8.

In the above description, the resource scheduler determines whether to perform resource scheduling based on the remaining tolerance of the application, or based on the remaining tolerance of the application and the system entropy. In actual applications, the resource scheduler can also determine whether to perform resource scheduling based on other methods. This situation is explained below.

The resource scheduler can obtain the first scheduling information or the second scheduling information from the service quality predictor, the first scheduling information indicates to increase the isolation area resources of the target LC type application among the multiple LC type applications, and the second scheduling information indicates to reduce the isolation area resources of the target LC type application. Then, according to the first scheduling information, the isolation area resources of the target LC type application are increased; or according to the second scheduling information, the isolation area resources of the target LC type application are reduced.

Specifically, if it is determined to increase the isolation area resources of the target LC type application, then the resource scheduler can determine that the isolation area of the target LC type application is a beneficiary area, based on step 601 of the embodiment shown in FIG6, step 603 to step 607, determine the victim area of this resource scheduling, and judge whether all can be successfully scheduled, and the details are not repeated here. If it is determined to reduce the isolation area resources of the target LC type application, then the resource scheduler can determine that the isolation area of the target LC type application is a victim area, based on step 602 to step 607 of the embodiment shown in FIG6, determine the beneficiary area of this resource scheduling, and judge whether all can be successfully scheduled, and the details are not repeated here.

In the embodiment of the present application, there are multiple ways to trigger the resource scheduler to perform resource scheduling. In addition to the remaining tolerance trigger, the remaining tolerance and the current system entropy trigger can also be used. In addition, it can also be triggered according to the indication information, which can adapt to different needs in actual applications and improve the flexibility and adaptability of the technical solution. In addition, the resource scheduler can directly determine the resource scheduling area based on the scheduling information, so that even if the resource scheduler cannot obtain the remaining tolerance or the current system entropy, it can still perform resource scheduling, ensuring the smooth operation of the resource scheduler and improving the reliability of the technical solution.

In some optional embodiments, the resource scheduler provided in the embodiments of the present application may also implement the following functions:

1) In the case where the computer system only includes LC type applications, calculate the current system entropy.

Assume that the computer system includes M BE type applications, where M is a positive integer. The resource scheduler can determine the system entropy E _BE of the BE type application according to the above formula 6. In this case, E _BE represents the amount of interference that the BE type application cannot tolerate, that is, the system entropy of the computer system with only BE type applications.

2) Based on Formula 8, determine the total anti-interference capability A of the computer system:

Formula 8:

The total anti-interference capability of the system A refers to the total anti-interference capability of the system including only N LC-type applications in the computer system, _{and Ai} represents the amount of interference that the user i can tolerate, which is explained in the above formula 1 and will not be repeated here.

3) Based on Formula 9, determine the total interference R to the computer system:

Formula 9:

The total interference amount R received by the system refers to the total interference amount received by the computer system including only N LC-type applications, and _Ri represents the actual interference amount received by user i, which is explained in the above formula 2 and will not be repeated here.

4) Based on Formula 10, determine the residual interference capacity S of the computer system:

Formula 10:

The system's residual interference capability S refers to the system's residual interference capability S of only N LC-type applications in the computer system, and _Si represents the residual interference tolerance of LC-type application i, which is explained in Formula 3 above and will not be repeated here.

In the embodiment of the present application, based on Formulas 1 to 10, the resource scheduler can evaluate the performance of the computer system from multiple perspectives.

In an embodiment of the present application, a resource scheduling method is also provided, which is applied to a service quality predictor.

The main principle of the service quality predictor is that the high tail latency of LC-type applications is caused by the queuing of requests. LC-type applications will queue requests from clients in a queue waiting for processing by the server. The latency of the request is composed of the system processing time and the queuing time. When the system has insufficient available resources or improper scheduling, the interference between applications will become serious, and it will take longer to process a request. Correspondingly, the average queue in the system will also become longer, increasing the queuing time of requests.

According to the conclusions in queuing theory, the growth of queue length will lead to an exponential increase in delay. The service quality predictor is based on the queuing phenomenon of delay-sensitive applications, which indirectly reflects the severity of interference to LC-type applications. Requests sent by clients may queue at multiple locations in the network, such as the network routing of the transit node, the network card buffer, the operating system buffer, etc.

The severity of the interference to the application directly affects the speed at which the application obtains data from the operating system buffer. The service quality predictor monitors the queues in the buffer corresponding to each application, thereby reflecting the degree of interference between different applications running on the same physical machine. The queues in the buffer can be measured by the average network receive queue (Recv-Q) length indicator.

Please refer to FIG. 9 , which is a flow chart of a resource scheduling method provided in an embodiment of the present application, comprising the following steps:

901. Obtain multiple network receiving queue lengths corresponding to the target LC type application.

Taking the continuous sampling of 30 data points in the buffer corresponding to the target LC type application as an example, the service quality predictor continuously samples these 30 data points at a fixed time interval to obtain 30 network receiving queue lengths.

It should be noted that the number of sampled network receive queue lengths and the sampling time interval can be selected according to actual application needs and are not specifically limited here.

902. Calculate the average of multiple network receive queue lengths to obtain an average network receive queue length.

After obtaining 30 network receiving queue lengths, the average network receiving queue length is used to reflect the interference degree of the target LC type application.

903. Determine whether the network is greater than a length threshold, if so, execute step 904, if not, execute step 905.

If the average network receive queue length corresponding to the target LC type application is greater than the length threshold, it is considered that the target KC type application is seriously interfered with and resources are scarce, and resources need to be allocated to it. If the average network receive queue length corresponding to the target LC type application is less than or equal to the length threshold, it is necessary to further determine whether to strip the isolation area resources of the target LC type application.

904. Send first scheduling information to the resource scheduler, where the first scheduling information indicates to increase the isolation area resources of the target LC type application.

When the average network receiving queue length corresponding to the target LC type application is greater than the length threshold, the service quality predictor sends first scheduling information to the resource scheduler to instruct the resource scheduler to increase isolation area resources for the target LC type application.

905. Determine whether the proportion of the network receiving queue lengths with a value of 0 among the multiple network receiving queue lengths is greater than the proportion threshold. If so, execute step 906; if not, return to step 901.

When the average network receiving queue length corresponding to the target LC type application is less than or equal to the length threshold, the service quality predictor needs to further determine whether the proportion of the sampled data equal to 0 meets the condition, that is, to determine whether the proportion of the network receiving queue length with a value of 0 in the multiple network receiving queue lengths is greater than the proportion threshold. If it is greater than the proportion threshold, it is considered that the isolation area resources of the target LC type application are redundant and can be stripped. If it is not greater than, the isolation area resources of the target LC type application are not adjusted, and the network receiving queue length of the target LC type application continues to be monitored.

The ratio threshold may be determined by a service quality predictor according to a historical operating state of the computer system, or may be determined by other devices (eg, computing devices) in the computer system, which is not specifically limited herein.

Exemplarily, assuming that among the 30 network receiving queue lengths, 25 network receiving queue lengths have a value of 0 and the ratio threshold is 80%, the service quality predictor will execute step 906 .

906. Send second scheduling information to the resource scheduler, where the second scheduling information indicates reducing the isolation area resources of the target LC type application.

When the average network receive queue length corresponding to the target LC type application is less than or equal to the length threshold, and the proportion of the network receive queue length with a value of 0 in multiple network receive queue lengths is greater than the proportion threshold, the service quality predictor sends second scheduling information to the resource scheduler to instruct the resource scheduler to reduce the isolation area resources for the target LC type application.

In the embodiment of the present application, the service quality predictor monitors the buffer corresponding to each LC type application, and can determine whether resource scheduling is required without calculating system entropy or tolerance, etc., thereby simplifying the process and saving computing resources.

In an embodiment of the present application, an application identification method is also provided, which is applied to an application distinguisher.

The application distinguisher distinguishes whether the application running on the virtual machine is an LC application or a BE application based on the data of the network send/receive bandwidth ratio of the virtual machine. The network send/receive bandwidth ratio of the LC application remains constant under different loads, while the network send/receive bandwidth ratio of the BE application changes greatly over time.

For example, please refer to FIG. 10a and FIG. 10b. FIG. 10a takes the sending bandwidth ratio of the Redis application in the LC type application as an example, and FIG. 10b takes the sending bandwidth ratio of the Terasort application based on the Spark framework in the BE application as an example. As shown in FIG. 10a, the change range of the sending bandwidth ratio of the Redis application is very small and can be considered to be constant. As shown in FIG. 10b, the change range of the sending bandwidth ratio of the Terasort application based on the Spark framework is very large.

It should be noted that both Figure 10a and Figure 10b use the sending bandwidth ratio as an example, which represents the ratio of the application's sending network bandwidth to the total bandwidth. In actual applications, the receiving bandwidth ratio can also be compared, where the receiving bandwidth ratio = 1 - the sending bandwidth ratio.

The reason for the phenomenon in Figures 10a and 10b is that LC applications and BE applications present different patterns in network data usage. Although the size of the network request packet sent/received by the same virtual machine may be different each time, it can be considered that the size of the network request packet sent/received always follows a certain distribution. Therefore, when a certain number of request data of the virtual machine in the same scenario are counted, the mean of the request size sent/received by the application should be close to the mean of the distribution. For these two different types of applications, LC applications can process more requests per unit time (usually 102-106), so the mean value obtained by multiple samplings when sampling LC applications should be basically constant; while BE applications can usually only process fewer requests per unit time (usually 0-10), and the sampled data usually varies greatly depending on which stage the BE application is in. Therefore, it can be considered that the ratio of the sending bandwidth to the receiving bandwidth of the LC application measured at different times changes little, while the ratio of the sending bandwidth to the receiving bandwidth of the BE application changes greatly.

Please refer to FIG. 11 , which is a flow chart of an application identification method provided in an embodiment of the present application, comprising the following steps:

1101. Obtain the average values of multiple network total bandwidths of multiple applications in the computer system at the current stage.

For example, take the average value of the total network bandwidth of an application as an example. Define the current stage as the i-th stage, _Xij represents the data of the j-th interval in the current stage, Represents the average value of the total network bandwidth at the current stage. The average value of the total network bandwidth at the current stage is determined based on Formula 11

Formula 11:

According to Formula 11, the average values of the total network bandwidths of multiple applications in the computing system at the current stage can be calculated.

In some optional embodiments, the computing device may monitor the total network bandwidth of each application, calculate the average value of the total network bandwidth, and send it to the application differentiator, so that the application differentiator obtains multiple average values of the total network bandwidth, which is not specifically limited here.

1102. Determine the coefficient of variation of multiple network total bandwidth intervals of multiple applications at the current stage based on the average values of multiple network total bandwidths.

After obtaining the average values of the total bandwidths of multiple networks, the application distinguisher can determine the interval variation coefficients ICOV of the total bandwidths of multiple networks based on Formula 12:

Formula 12:

In some optional embodiments, the computing device may calculate the interval variation coefficient of the total network bandwidth of each application and send it to the application differentiator, so that the application differentiator obtains the interval variation coefficient of the total network bandwidth, which is not specifically limited here.

1103. If the interval variation coefficients of the multiple network total bandwidths are greater than the interval coefficient threshold, then obtain the absolute values of the differences between the multiple end send/receive bandwidth ratios of the multiple applications in the current stage and the multiple start send/receive bandwidth ratios in the next stage.

The application distinguisher needs to monitor the stage changes of program operation. This is because BE-type applications may be in the process of network data transmission or calculation for a long time, which will cause the collected data to remain constant and lead to misjudgment. Therefore, it is necessary to detect the stage changes of program operation and analyze the data before and after the stage changes.

Therefore, the application distinguisher will determine whether the interval variation coefficient of the total network bandwidth is greater than the interval coefficient threshold. If the interval variation coefficient of the total network bandwidth of a certain application is greater than the interval coefficient threshold, it means that the application has begun to undergo a phased change, and the data before and after the phase change of the application will be collected and analyzed. Among them, the interval coefficient threshold can be set according to the needs of the actual application. The smaller the interval coefficient threshold, the more subtle phase changes can be detected, so that more phases are detected; the larger the interval coefficient threshold, the larger the phase changes of the application can be detected. Exemplarily, the interval coefficient threshold can be 0.25.

When the coefficient of variation of the total network bandwidth interval is greater than the interval coefficient threshold, the application distinguisher obtains the absolute value of the difference between the send/receive bandwidth ratio of the application at the end of the current stage and the send/receive bandwidth ratio at the beginning of the next stage. The send bandwidth ratio refers to the ratio of the application's send network bandwidth to the total network bandwidth; the receive bandwidth ratio refers to the ratio of the application's receive network bandwidth to the total network bandwidth.

For example, taking the transmission bandwidth ratio as an example, assuming that 6 data are collected in the current stage, the transmission bandwidth ratio of the last data among the 6 data can be determined as the end transmission bandwidth ratio of the current stage. The next data collected after the current stage is used as the starting collection data of the next stage, and the transmission bandwidth ratio of the data is used as the starting transmission bandwidth ratio of the next stage.

In some optional embodiments, a computing device may calculate multiple absolute values of the differences and send the values to the application distinguisher, so that the application distinguisher obtains the multiple absolute values of the differences, which is not specifically limited here.

1104. Obtain multiple sending/receiving bandwidth ratio variation coefficients of multiple applications in the target time periods before and after the current stage.

For example, take the coefficient of variation of the send/receive bandwidth ratio of an application as an example. Assuming the target time period is 10 seconds, the coefficient of variation COV of the send/receive data ratio during the period from the first 10 seconds of the current stage to the last 10 seconds of the current stage is calculated. Among them, COV is the ratio of the standard deviation σ of the send/receive bandwidth ratio during this period to the average value μ of the send/receive bandwidth ratio during this period, which is a normalized measure used to measure the discrete degree of data probability distribution.

In some optional embodiments, a computing device may calculate multiple transmission/reception bandwidth ratio variation coefficients and send them to an application differentiator, so that the application differentiator obtains multiple transmission/reception bandwidth ratio variation coefficients, which are not specifically limited here.

1105. According to multiple absolute values of differences or multiple transmission/reception bandwidth ratio variation coefficients, determine that the identifier of the LC type application in the multiple applications is the first application identifier, and the identifier of the BE type application in the multiple applications is the second application identifier.

The application distinguisher can determine, from multiple applications, that an application whose absolute value of difference is greater than a difference threshold and/or whose coefficient of variation of the transmission/reception bandwidth ratio is greater than a coefficient threshold is a BE type application. And determine that applications other than the BE type applications in the multiple applications are LC type applications. Then, the identifier of the LC type application is marked as a first application identifier, and the identifier of the BE type application is marked as a second application identifier.

Specifically, there are various bases for the application distinguisher to distinguish application types, and possible situations are described below respectively.

1) Determine among multiple applications that the absolute value of the difference is greater than the difference threshold as a BE type application, and the other applications are LC type applications.

2) Determine among multiple applications that the transmission/reception bandwidth ratio variation coefficient is greater than a coefficient threshold as a BE type application, and the other applications are LC type applications.

3) Determine that the applications whose absolute values of differences are greater than a difference threshold and whose coefficient of variation of the transmission/reception bandwidth ratio is greater than a coefficient threshold are BE type applications, and the other applications are LC type applications.

In the embodiment of the present application, there are multiple cases for determining the application type. The above case 1) or case 2) does not require a very high distinguishing function of the application distinguisher, allowing a certain error tolerance. Case 3) is relatively strict, which is conducive to clearly distinguishing LC type applications from BE type applications, and improves the reliability of the system.

1106. Send the first application identifier and the second application identifier to the resource scheduler so that the resource scheduler can distinguish between BE type applications and LC type applications.

After identifying the first application identifier and the second application identifier, the application distinguisher may send the first application identifier and the second application identifier to the resource scheduler, so that the resource scheduler determines the LC type application according to the first application identifier and determines the BE type application according to the second application identifier.

It should be noted that in the embodiments of the present application, the interval coefficient threshold, the coefficient threshold, and the difference threshold can all be determined according to the needs of the actual application, and are not specifically limited here.

In an embodiment of the present application, the application distinguisher can identify the application types of multiple applications in a computer system and inform the resource scheduler of the application identifier of each application, providing technical support for the resource scheduler to perform resource scheduling and improving the feasibility of the technical solution.

Through the description of the embodiment shown in FIG11, it can be known that the application distinguisher actually performs three judgments and performs different operations according to different judgment results. Taking the embodiment shown in FIG11, the application whose absolute value of difference is greater than the difference threshold or the application whose transmission bandwidth ratio variation coefficient is greater than the coefficient threshold is determined as a BE type application and the other applications are LC type applications as an example, the method of application identification is described.

Please refer to FIG. 12 , which is a flow chart of an application identification method provided in an embodiment of the present application, comprising the following steps:

1201. Detect phase changes of target applications.

The application distinguisher detects the phase change of the target application by detecting the coefficient of variation of the total network bandwidth interval of the target application in the current phase. Specifically, it is similar to step 1101 and step 1102 in the embodiment shown in FIG11 , and will not be described in detail here.

1202. Determine whether the target application undergoes a phased change. If so, execute step 1203; if not, execute step 1201.

If the interval variation coefficient of the total network bandwidth of the target application in the current stage is greater than the interval coefficient threshold, it is considered that the target application has undergone a phased change; if the interval variation coefficient of the total network bandwidth of the target application in the current stage is not greater than the interval coefficient threshold, it is considered that the target application has not undergone a phased change.

1203. Determine whether the change in the sending bandwidth ratio meets the conditions, if so, execute step 1206, if not, execute step 1204.

When it is determined that the target application has undergone a phased change, the application distinguisher will further determine whether the change in the target application's transmission bandwidth ratio meets the condition, that is, whether the absolute value of the difference between the target application's transmission bandwidth ratio at the end of the current phase and the starting transmission bandwidth ratio of the next phase is greater than the difference threshold. If so, the condition is considered to be met, otherwise, the condition is considered not to be met.

1204. Determine whether the transmission bandwidth ratio variation coefficient is greater than the coefficient threshold, if so, execute step 1206, if not, execute step 1205.

When the change in the sending bandwidth ratio does not meet the condition, the application distinguisher determines whether the sending bandwidth ratio variation coefficient in the target time period before and after the current application stage is greater than the coefficient threshold, and performs different operations according to different results.

1205. Determine that the target application is an LC type application and mark the first application identifier.

1206. Determine that the target application is an LC type application and mark the second application identifier.

Step 1205 and step 1206 are described in step 1105 in the embodiment shown in FIG. 11 , and will not be described again here.

The embodiment of the present application further provides a resource scheduling system, please refer to Figure 13a, which is a schematic diagram of the structure of the resource scheduling system provided by the embodiment of the present application. App1 to AppN in Figure 13a represent N applications.

As shown in FIG13a, the resource scheduling system includes a resource scheduler, and the resource scheduler is used to: obtain the residual interference tolerance of each LC type application in at least one delay-sensitive LC type application included in the computer system. From multiple LC type applications, obtain a first LC type application with the smallest residual interference tolerance. If the residual interference tolerance of the first LC type application is less than the lower limit of the tolerance, increase the first isolation area resources of the first LC type application. If the residual interference tolerance of the first LC type application is greater than the upper limit of the tolerance, transfer the second isolation area resources of the second LC type application in the multiple LC type applications to the resource sharing area.

The resource scheduler may also perform the operations performed by the resource scheduler in the embodiments shown in the aforementioned FIG. 1a to FIG. 8 , which will not be described in detail here.

In some optional embodiments, the resource scheduling system further includes a service quality predictor, which is used to: obtain multiple network receive queue lengths corresponding to the target LC type application. Calculate the average of the multiple network receive queue lengths to obtain the average network receive queue length. If the average network receive queue length is greater than the length threshold, send first scheduling information to the resource scheduler, and the first scheduling information indicates to increase the isolation area resources of the target LC type application. If the average network receive queue length is less than or equal to the length threshold, and the proportion of the network receive queue length with a value of 0 in the multiple network receive queue lengths is greater than the proportion threshold, send second scheduling information to the resource scheduler, and the second scheduling information indicates to reduce the isolation area resources of the target LC type application.

The service quality predictor may also perform the operations performed by the service quality predictor in the embodiment shown in FIG. 9 , which will not be described in detail here.

In some optional embodiments, the resource scheduling system further includes an application distinguisher, which is used to: obtain multiple average network total bandwidths of multiple applications in the computer system at the current stage. According to the multiple average network total bandwidths, determine the multiple interval variation coefficients of the multiple network total bandwidths of the multiple applications at the current stage. If the multiple interval variation coefficients of the multiple network total bandwidths are greater than the interval coefficient threshold, obtain multiple absolute values of the difference between the multiple end send/receive bandwidth ratios of the multiple applications at the current stage and the multiple start send/receive bandwidth ratios of the next stage. Obtain multiple coefficients of variation of the send/receive bandwidth ratios of the multiple applications in the target time periods before and after the current stage. According to the multiple absolute values of the difference or the multiple coefficients of variation of the send/receive bandwidth ratios, determine that the identifier of the LC type application in the multiple applications is the first application identifier, and the identifier of the BE type application in the multiple applications is the second application identifier. Send the first application identifier and the second application identifier to the resource scheduler so that the resource scheduler distinguishes between the BE type application and the LC type application.

The application distinguisher can also perform the operations performed by the application distinguisher in the embodiments shown in the above-mentioned Figures 11 to 12, which will not be repeated here.

In some optional embodiments, the resource scheduling system also includes a computing device, which is used to send to the resource scheduler at least one of the current system entropy, or the amount of interference that each LC-type application can tolerate, or the amount of interference actually received by each LC-type application, or the first number of instructions per cycle when each BE-type application runs alone, or the second number of instructions per cycle after each BE-type application is interfered.

In some optional embodiments, the computing device is further configured to obtain a plurality of network receiving queue lengths corresponding to the target LC type application, calculate the average of the plurality of network receiving queue lengths to obtain an average network receiving queue length, and send the average network receiving queue length to the service quality predictor.

In some optional embodiments, the computing device is also used to send to the application distinguisher an average value of multiple network total bandwidths of multiple applications in the computer system at the current stage, or multiple network total bandwidth interval variation coefficients of multiple applications at the current stage, or multiple absolute values of differences between multiple end send/receive bandwidth ratios of multiple applications at the current stage and multiple start send/receive bandwidth ratios of the next stage, or at least one of multiple send/receive bandwidth ratio variation coefficients of multiple applications in the target time periods before and after the current stage.

It is understandable that if the resource scheduling system does not include a separate computing device, the resource scheduler may include a computing module for calculating data related to the current system entropy, and the remaining tolerance, etc., which are not specifically limited here. The service quality predictor can calculate and determine the average network receive queue length by itself. The application distinguisher can calculate and determine data related to the total network bandwidth and the send/receive bandwidth data related to each application by itself, which are not specifically limited here.

The resource scheduling system provided in the embodiment of the present application can be applied in a "black box scenario", please refer to Figure 13b, which is another structural diagram of the resource scheduling system provided in the embodiment of the present application.

In scenarios such as public clouds, due to requirements for user data security and privacy, cloud service providers are usually unable to obtain application information running inside virtual machines (such as application name, application type, application tail latency, etc.). They can only collect data such as the status of physical resources used by the virtual machines and system overhead at the operating system level. This scenario is called a "black box scenario."

In FIG. 13 b , the functions of the various components are similar to those of the various components in FIG. 13 a , and are not described in detail here.

The embodiment of the present application also provides a resource scheduler, which is applied to a cloud computing system. Please refer to Figure 14, which is a schematic diagram of the structure of the resource scheduler provided in the embodiment of the present application.

As shown in FIG. 14 , the resource scheduler 1400 includes an acquisition unit 1401 and a processing unit 1402 .

The acquisition unit 1401 is used to acquire the residual interference tolerance of each LC type application in at least one delay-sensitive LC type application included in the computer system; and acquire the first LC type application with the smallest residual interference tolerance from multiple LC type applications.

The processing unit 1402 is used to:

If the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, the first isolation area resources of the first LC type application are increased.

If the remaining interference tolerance of the first LC-type application is greater than the tolerance upper limit, the second isolation area resources of the second LC-type application among the multiple LC-type applications are transferred to the resource sharing area.

In some optional embodiments, the acquisition unit 1401 is further used to acquire the current system entropy of the computer system, where the current system entropy is used to indicate the degree of interference between applications in the current computer system.

The processing unit 1402 is specifically configured to:

If the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, then the resources of the first isolation area are increased.

If the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is greater than the tolerance upper limit, the second isolation area resources of the second LC type application are transferred to the resource sharing area.

In some optional embodiments, the processing unit 1402 is specifically configured to:

If there is a third LC type application in the computer system whose residual interference tolerance is greater than the tolerance upper limit and which has a third isolation area resource that can be stripped, then the resources in the third isolation area resource are transferred to the first isolation area corresponding to the first LC type application to increase the first isolation area resources;

If the third LC type application does not exist in the computer system, the resources of the resource sharing area are transferred to the first isolation area to increase the resources of the first isolation area.

In some optional embodiments, the acquiring unit 1401 is specifically configured to:

Obtain the amount of interference that each LC type application can tolerate and the amount of interference each LC type application actually receives.

The current system entropy is determined based on the amount of interference that each LC type application can tolerate and the amount of interference that each LC type application actually receives.

In some optional embodiments, the computer system further includes at least one best effort BE type application.

The acquisition unit 1401 is further used to acquire a first application identifier and a second application identifier from an application distinguisher or from a user, wherein the first application identifier is used to indicate an LC type application, and the second application identifier is used to indicate a BE type application.

The processing unit 1402 is further configured to determine an LC type application in the computer system according to the first application identifier; and to determine a BE type application in the computer system according to the second application identifier.

The acquisition unit 1401 is specifically used for:

Obtain the amount of interference that each LC type application can tolerate and the amount of interference each LC type application actually receives.

The entropies of the plurality of LC type applications are determined according to the amount of interference that each LC type application can tolerate and the amount of interference that each LC type application actually receives.

A first number of instructions per cycle when each BE type application in at least one BE type application runs independently and a second number of instructions per cycle after each BE type application is disturbed are obtained.

An entropy of at least one BE-type application is determined based on the first number of instructions per cycle and the second number of instructions per cycle.

A current system entropy is determined based on the entropies of the plurality of LC-type applications and the entropy of at least one BE-type application.

In some optional embodiments, the acquisition unit 1401 is also used to obtain first scheduling information or second scheduling information from a service quality predictor, the first scheduling information indicates increasing the isolation area resources of a target LC type application among multiple LC type applications, and the second scheduling information indicates reducing the isolation area resources of the target LC type application.

The processing unit 1402 is further configured to increase the isolation area resources of the target LC type application according to the first scheduling information; or to reduce the isolation area resources of the target LC type application according to the second scheduling information.

In some optional embodiments, the acquisition unit 1401 is further used to acquire system entropy after computer system resources are scheduled.

The processing unit 1402 is further configured to determine that the resource scheduling is successful if the system entropy after the resource scheduling is less than the current system entropy.

The resource scheduler 1400 can execute the operations performed by the resource scheduler in the embodiments shown in the aforementioned Figures 1a to 8, and Figures 13a and 13b, which will not be repeated here.

The embodiment of the present application also provides a service quality predictor, which is applied to a cloud computing system. Please refer to Figure 15, which is a schematic diagram of the structure of the service quality predictor provided by the embodiment of the present application.

As shown in FIG. 15 , the service quality predictor includes an acquisition unit 1501 , a processing unit 1502 and a sending unit 1503 .

The acquisition unit 1501 is used to acquire multiple network receiving queue lengths corresponding to the target LC type application.

The processing unit 1502 is used to calculate the average of the lengths of multiple network receiving queues to obtain an average network receiving queue length;

The sending unit 1503 is used to:

If the average network receiving queue length is greater than the length threshold, first scheduling information is sent to the resource scheduler, where the first scheduling information indicates to increase the isolation area resources of the target LC type application;

If the average network receive queue length is less than or equal to the length threshold, and the proportion of network receive queue lengths with a value of 0 among multiple network receive queue lengths is greater than the proportion threshold, then a second scheduling information is sent to the resource scheduler, and the second scheduling information indicates to reduce the isolation area resources of the target LC type application.

The service quality predictor 1500 can execute the operations executed by the service quality predictor in the embodiments shown in the aforementioned FIG. 9 , FIG. 13 a and FIG. 13 b , which will not be described in detail here.

The embodiment of the present application also provides an application distinguisher. Please refer to Figure 16, which is a structural diagram of the application distinguisher provided in the embodiment of the present application.

As shown in FIG. 16 , the service quality predictor includes an acquisition unit 1601 , a processing unit 1602 and a sending unit 1603 .

The acquisition unit 1601 is used to acquire the average values of the total network bandwidths of multiple applications in the computer system at the current stage.

The processing unit 1602 is used to determine multiple interval variation coefficients of the total network bandwidths of multiple applications at the current stage according to the average values of the multiple total network bandwidths.

The acquisition unit 1601 is further configured to acquire, if the multiple network total bandwidth interval variation coefficients are greater than the interval coefficient threshold, multiple absolute values of differences between multiple end send/receive bandwidth ratios of multiple applications in the current phase and multiple start send/receive bandwidth ratios of multiple applications in the next phase. Multiple send/receive bandwidth ratio variation coefficients of multiple applications in the target time periods before and after the current phase.

The processing unit 1602 is further configured to determine, based on multiple difference absolute values or multiple transmission/reception bandwidth ratio variation coefficients, an identifier of an LC type application among multiple applications as a first application identifier and an identifier of a BE type application among multiple applications as a second application identifier.

The sending unit 1603 is used to send the first application identifier and the second application identifier to the resource scheduler, so that the resource scheduler can distinguish between the BE type application and the LC type application.

In some optional embodiments, the processing unit 1602 is specifically configured to:

From among multiple applications, applications whose absolute values of differences are greater than a difference threshold and/or whose coefficient of variation of the transmission/reception bandwidth ratio is greater than a coefficient threshold are determined to be BE type applications.

The identifier for marking the BE type application is the second application identifier.

An identifier of an application other than the BE type application among the multiple applications is marked as a first application identifier.

The application distinguisher 1600 can perform the operations performed by the application distinguisher in the embodiments shown in the aforementioned Figures 11 to 13b, which will not be repeated here.

Next, the computer device provided in the embodiment of the present application is described, and please refer to Figure 17, which is a schematic diagram of the structure of the computer device provided in the embodiment of the present application. The computer device 1700 includes: a processor 1701 and a memory 1702, and the memory 1702 stores one or more application programs or data.

Among them, the memory 1702 can be a volatile storage or a persistent storage. The program stored in the memory 1702 can include one or more modules, each of which can be used to execute a series of operations performed by the computer device 1700. Furthermore, the processor 1701 can communicate with the memory 1702 and execute a series of instruction operations in the memory 1702 on the computer device 1700. The processor 1701 can be a central processing unit (CPU) or a single-core processor. In addition, it can also be other types of processors, such as a dual-core processor, which is not limited here.

The computer device 1700 may further include one or more communication interfaces 1703 , one or more operating systems, such as Windows Server ^™ , Mac OS X ^™ , Unix ^™ , Linux ^™ , FreeBSD ^™ , etc.

In some optional embodiments, the computer device may further include a last level cache (LLC) and a main memory, which are not specifically limited here.

When the computer device 1700 is used as a resource scheduler, it can execute the operations performed by the resource scheduler in the embodiments shown in Figures 1a to 8, 13a and 13b; when the computer device 1700 is used as a service quality predictor, it can execute the operations performed by the service quality predictor in the embodiments shown in Figures 1a to 9, 13a and 13b; when the computer device 1700 is used as an application differentiator, it can execute the operations performed by the application differentiator in the embodiments shown in Figures 1a to 8, 11 to 13b; no further details will be given here.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

Claims (28)

1. A resource scheduling method for a cloud computing system, characterized in that the method is applied to a resource scheduler, and the method includes: Obtaining the remaining interference tolerance of each of the plurality of delay-sensitive LC-type applications included in the computer system; From the plurality of LC-type applications, obtain the first LC-type application with the smallest remaining interference tolerance; If the remaining interference tolerance of the first LC-type application is less than the lower limit of tolerance, increase the first isolation area resources of the first LC-type application; If the remaining interference tolerance of the first LC-type application is greater than the tolerance upper limit, the second isolation area resources of the second LC-type application among the plurality of LC-type applications are transferred to the resource sharing area. 2. The method according to claim 1, characterized in that, the method further comprises: Obtaining current system entropy of the computer system, where the current system entropy is used to currently indicate the degree of interference between applications in the computer system; If the remaining interference tolerance of the first LC type application is less than the lower limit of tolerance, increasing the first isolation area resources of the first LC type application includes: If the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, increase the first isolation zone resources; If the remaining interference tolerance of the first LC-type application is greater than the tolerance upper limit, then transferring the second isolation area resources of the second LC-type application among the plurality of LC-type applications to the resource sharing area includes: If the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC-type application is greater than the tolerance upper limit, then the second isolation zone resources of the second LC-type application are Move to resource sharing area. 3. The method according to claim 1 or 2, characterized in that said increasing the isolation zone resources of the first LC type application includes: If there is a third LC-type application in the computer system whose remaining interference tolerance is greater than the tolerance upper limit and has a third isolation area resource that can be stripped, then the resources in the third isolation area resource are transferred to the third isolation area resource. The first isolation area corresponding to the first LC type application to increase the resources of the first isolation area; If the third LC type application does not exist in the computer system, the resources in the resource sharing area are transferred to the first isolation area to increase the resources in the first isolation area. 4. The method according to claim 2 or 3, characterized in that said obtaining the current system entropy of the computer system includes: Obtain the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives; The current system entropy is determined based on the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives. 5. The method according to claim 2 or 3, wherein the computer system further includes at least one best-effort BE application; before obtaining the current system entropy of the computer system, the method further includes: Obtain the first application identifier and the second application identifier from the application distinguisher or from the user, the first application identifier is used to indicate the LC type application, and the second application identifier is used to indicate the BE type application; Determine the LC type application in the computer system according to the first application identifier; Determine the BE application in the computer system according to the second application identifier; The method of obtaining the current system entropy of the computer system includes: Obtain the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives; Determine the entropy of the plurality of LC-type applications based on the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives; Obtain the first number of instructions per cycle when each BE type application in the at least one BE type application is running alone and the second number of instructions per cycle after each BE type application is disturbed; Determine the entropy of the at least one BE-type application based on the first number of instructions per cycle and the second number of instructions per cycle; The current system entropy is determined based on the entropy of the plurality of LC-type applications and the entropy of the at least one BE-type application. 6. The method according to any one of claims 1 to 5, characterized in that the method further comprises: Obtain first scheduling information or second scheduling information from the quality of service predictor, the first scheduling information indicates increasing isolation area resources of the target LC type application among the plurality of LC type applications, and the second scheduling information indicates Reduce the isolation area resources of the target LC type application; According to the first scheduling information, increase the isolation area resources of the target LC type application; Alternatively, according to the second scheduling information, the isolation area resources of the target LC type application are reduced. 7. The method according to any one of claims 2 to 6, characterized in that the method further comprises: Obtain the system entropy after resource scheduling of the computer system; If the system entropy after resource scheduling is less than the current system entropy, it is determined that resource scheduling is successful. 8. A resource scheduling method for a cloud computing system, characterized in that the method is applied to a quality of service predictor, and the method includes: Obtain the lengths of multiple network receive queues corresponding to the target LC type application; Calculate the average value of the multiple network reception queue lengths to obtain the average network reception queue length; If the average network reception queue length is greater than the length threshold, send first scheduling information to the resource scheduler, where the first scheduling information indicates adding isolation area resources for the target LC type application; If the average network receive queue length is less than or equal to the length threshold, and the network receive queue length with a value of 0 among the multiple network receive queue lengths accounts for greater than the proportion of the multiple network receive queue lengths, threshold, the second scheduling information is sent to the resource scheduler, and the second scheduling information indicates reducing the isolation area resources of the target LC type application. 9. An application identification method, characterized in that the method is applied to an application distinguisher, and the method includes: Obtain the average total network bandwidth of multiple applications in the computer system at the current stage; Determine the coefficients of variation of multiple total network bandwidth intervals of the multiple applications in the current stage based on the average values of the multiple total network bandwidths; If the interval variation coefficient of the multiple total network bandwidths is greater than the interval coefficient threshold, obtain the multiple end sending/receiving bandwidth ratios of the multiple applications in the current stage and the multiple starting sending/receiving bandwidths of the next stage. The absolute values of multiple differences in proportions; Obtain multiple transmit/receive bandwidth ratio variation coefficients of the multiple applications within the target time periods before and after the current stage; According to the plurality of absolute difference values or the plurality of transmit/receive bandwidth ratio variation coefficients, the identifier of the LC-type application among the plurality of applications is determined to be the first application identifier, and the identifier of the BE-type application among the plurality of applications is determined The identifier is the second application identifier; The first application identifier and the second application identifier are sent to a resource scheduler, so that the resource scheduler distinguishes the BE type application and the LC type application. 10. The method according to claim 9, characterized in that, based on the plurality of absolute difference values or the plurality of transmission/reception bandwidth ratio variation coefficients, the identification of the LC type application in the plurality of applications is determined to be The first application identifier, the identifier of the BE application among the multiple applications is the second application identifier, including: From the multiple applications, it is determined that the absolute value of the difference is greater than the difference threshold, and/or the application whose transmission/reception bandwidth ratio variation coefficient is greater than the coefficient threshold is a BE-type application; The identifier marking the BE-type application is the second application identifier; An identifier marking applications other than the BE-type application among the plurality of applications is a first application identifier. 11. A resource scheduling system, characterized in that the resource scheduling system includes a resource scheduler, and the resource scheduler is used for: Obtaining a residual interference tolerance for each of the at least one delay-sensitive LC-type application included in the computer system; From the plurality of LC-type applications, obtain the first LC-type application with the smallest remaining interference tolerance; If the remaining interference tolerance of the first LC-type application is less than the lower limit of tolerance, increase the first isolation area resources of the first LC-type application; If the remaining interference tolerance of the first LC-type application is greater than the tolerance upper limit, the second isolation area resources of the second LC-type application among the plurality of LC-type applications are transferred to the resource sharing area. 12. The resource scheduling system according to claim 11, characterized in that the resource scheduler is also used to: Obtain the first application identifier and the second application identifier from the application distinguisher, the first application identifier is used to indicate the LC type application, and the second application identifier is used to indicate the BE type application; Determine the LC type application in the computer system according to the first application identifier; According to the second application identifier, a BE-type application in the computer system is determined. 13. The resource scheduling system according to claim 12, characterized in that the resource scheduling system further includes the application differentiator, the application differentiator is used for: Obtain the average total network bandwidth of multiple applications in the computer system at the current stage; Determine the coefficients of variation of multiple total network bandwidth intervals of the multiple applications in the current stage based on the average values of the multiple total network bandwidths; If the interval variation coefficient of the multiple total network bandwidths is greater than the interval coefficient threshold, obtain the multiple end sending/receiving bandwidth ratios of the multiple applications in the current stage and the multiple starting sending/receiving bandwidths of the next stage. The absolute values of multiple differences in proportions; Obtain multiple transmit/receive bandwidth ratio variation coefficients of the multiple applications within the target time periods before and after the current stage; According to the plurality of absolute difference values or the plurality of transmit/receive bandwidth ratio variation coefficients, the identifier of the LC-type application among the plurality of applications is determined to be the first application identifier, and the identifier of the BE-type application among the plurality of applications is determined The identifier is the second application identifier; The first application identifier and the second application identifier are sent to the resource scheduler, so that the resource scheduler distinguishes the BE type application and the LC type application. 14. The resource scheduling system according to any one of claims 11 to 13, characterized in that the resource scheduler is also used to: Obtain first scheduling information or second scheduling information from the quality of service predictor, the first scheduling information indicates increasing isolation area resources of the target LC type application among the plurality of LC type applications, and the second scheduling information indicates Reduce the isolation area resources of the target LC type application; According to the first scheduling information, increase the isolation area resources of the target LC type application; Alternatively, according to the second scheduling information, the isolation area resources of the target LC type application are reduced. 15. The resource scheduling system according to claim 14, characterized in that the resource scheduling system further includes the service quality predictor, and the service quality predictor is used for: Obtain the lengths of multiple network reception queues corresponding to the target LC type application; Calculate the average value of the multiple network reception queue lengths to obtain the average network reception queue length; If the average network reception queue length is greater than the length threshold, send first scheduling information to the resource scheduler, and the first scheduling information indicates increasing the isolation area resources of the target LC type application; If the average network receive queue length is less than or equal to the length threshold, and the network receive queue length with a value of 0 among the multiple network receive queue lengths accounts for greater than the proportion of the multiple network receive queue lengths, threshold, the second scheduling information is sent to the resource scheduler, and the second scheduling information indicates reducing the isolation area resources of the target LC type application. 16. A resource scheduler, characterized in that the resource scheduler is applied to a cloud computing system, and the resource scheduler includes: an acquisition unit configured to acquire the remaining interference tolerance of each of the at least one delay-sensitive LC-type application included in the computer system; The acquisition unit is further configured to acquire the first LC-type application with the smallest remaining interference tolerance from the plurality of LC-type applications; A processing unit configured to increase the first isolation zone resources of the first LC-type application if the remaining interference tolerance of the first LC-type application is less than the lower limit of tolerance; The processing unit is also configured to, if the remaining interference tolerance of the first LC-type application is greater than the tolerance upper limit, transfer the second isolation area resource of the second LC-type application among the plurality of LC-type applications to Resource sharing area. 17. The resource scheduler according to claim 16, characterized in that the obtaining unit is further used to obtain the current system entropy of the computer system, and the current system entropy is used to indicate the current relationship between applications in the computer system. degree of interference; The processing unit is specifically used for: If the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC type application is less than the tolerance lower limit, increase the first isolation zone resources; If the current system entropy is greater than or equal to the system entropy threshold, and the remaining interference tolerance of the first LC-type application is greater than the tolerance upper limit, then the second isolation zone resources of the second LC-type application are Move to resource sharing area. 18. The resource scheduler according to claim 16 or 17, characterized in that the processing unit is specifically used to: If there is a third LC-type application in the computer system whose remaining interference tolerance is greater than the tolerance upper limit and has a third isolation area resource that can be stripped, then the resources in the third isolation area resource are transferred to the third isolation area resource. The first isolation area corresponding to the first LC type application to increase the resources of the first isolation area; If the third LC type application does not exist in the computer system, the resources in the resource sharing area are transferred to the first isolation area to increase the resources in the first isolation area. 19. The resource scheduler according to claim 17 or 18, characterized in that the acquisition unit is specifically used to: Obtain the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives; The current system entropy is determined based on the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives. 20. The resource scheduler according to claim 17 or 18, characterized in that the computer system further includes at least one best-effort BE application; the acquisition unit is also used to: Obtain the first application identifier and the second application identifier from the application distinguisher or from the user, the first application identifier is used to indicate the LC type application, and the second application identifier is used to indicate the BE type application; Determine the LC type application in the computer system according to the first application identifier; Determine the BE application in the computer system according to the second application identifier; The acquisition unit is specifically used for: Obtain the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives; Determine the entropy of the plurality of LC-type applications based on the amount of interference that each LC-type application can tolerate and the amount of interference that each LC-type application actually receives; Obtain the first number of instructions per cycle when each BE type application in the at least one BE type application is running alone and the second number of instructions per cycle after each BE type application is disturbed; Determine the entropy of the at least one BE-type application based on the first number of instructions per cycle and the second number of instructions per cycle; The current system entropy is determined based on the entropy of the plurality of LC-type applications and the entropy of the at least one BE-type application. 21. The resource scheduler according to any one of claims 16 to 20, characterized in that the acquisition unit is also used to: Obtain first scheduling information or second scheduling information from the quality of service predictor, the first scheduling information indicates increasing isolation area resources of the target LC type application among the plurality of LC type applications, and the second scheduling information indicates Reduce the isolation area resources of the target LC type application; The processing unit is also used for: According to the first scheduling information, increase the isolation area resources of the target LC type application; Alternatively, according to the second scheduling information, the isolation area resources of the target LC type application are reduced. 22. The resource scheduler according to any one of claims 17 to 21, characterized in that the obtaining unit is also used to obtain the system entropy after resource scheduling of the computer system; The processing unit is also configured to determine that resource scheduling is successful if the system entropy after resource scheduling is less than the current system entropy. 23. A service quality predictor, characterized in that the service quality predictor is applied to a cloud computing system, and the service quality predictor includes: The acquisition unit is used to obtain the lengths of multiple network reception queues corresponding to the target LC type application; A processing unit configured to calculate the average of the multiple network reception queue lengths to obtain the average network reception queue length; A sending unit configured to send first scheduling information to a resource scheduler if the average network reception queue length is greater than a length threshold, where the first scheduling information indicates increasing the isolation area resources of the target LC type application; The sending unit is also configured to: if the average network reception queue length is less than or equal to the length threshold, and the network reception queue length with a value of 0 among the multiple network reception queue lengths is received in the multiple network If the proportion of the queue length is greater than the proportion threshold, second scheduling information is sent to the resource scheduler, and the second scheduling information indicates reducing the isolation area resources of the target LC type application. 24. An application differentiator, characterized by comprising: The acquisition unit is used to obtain the average total network bandwidth of multiple applications in the computer system at the current stage; A processing unit configured to determine the coefficients of variation of multiple total network bandwidth intervals of the multiple applications in the current stage based on the average values of the multiple total network bandwidths; The acquisition unit is also configured to acquire the sending/receiving bandwidth ratio of the multiple applications at multiple ends of the current phase and the ratio of the next phase if the interval variation coefficient of the multiple total network bandwidths is greater than the interval coefficient threshold. Multiple difference absolute values of multiple initial transmit/receive bandwidth ratios; The acquisition unit is also configured to acquire multiple transmit/receive bandwidth ratio variation coefficients of the multiple applications within the target time periods before and after the current stage; The processing unit is further configured to determine, based on the plurality of absolute difference values or the plurality of transmission/reception bandwidth ratio variation coefficients, the identification of the LC-type application in the plurality of applications as the first application identification, said The ID of the BE-type application in multiple applications is the second application ID; A sending unit, configured to send the first application identifier and the second application identifier to a resource scheduler, so that the resource scheduler distinguishes the BE type application and the LC type application. 25. The application differentiator according to claim 24, characterized in that the processing unit is specifically used for: From the multiple applications, it is determined that the absolute value of the difference is greater than the difference threshold, and/or the application whose transmission/reception bandwidth ratio variation coefficient is greater than the coefficient threshold is a BE-type application; The identifier marking the BE-type application is the second application identifier; An identifier marking applications other than the BE-type application among the plurality of applications is a first application identifier. 26. A computer device, characterized by comprising: a processor, a memory and a communication interface; The processor, the memory and the communication interface are connected; The processor is configured to perform the method according to any one of claims 1 to 10. 27. A computer-readable storage medium, characterized in that the computer-readable storage medium stores instructions that, when run on a computer, cause the computer to execute any one of claims 1 to 10. method described. 28. A computer program product, characterized in that, when the computer program product is executed on a computer, the computer performs the method of any one of claims 1 to 10.