JP2007265244A - Web system performance monitoring device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent the future overflow of performance by predicting the performance of a Web system. <P>SOLUTION: A data acquisition section 12 acquires a predetermined parameter from the Web system to be monitored, for providing service on a network. A model storage section 18 stores a base model used as a base of a prediction model for predicting the performance of the Web system. A factor input section 14 receives the input of a factor which is data influencing a predicted result using the prediction model. A prediction model preparing section 16 prepares the prediction model by correcting the base model using the factor. A performance overflow predicting section 22 determines a performance prediction value serving as data which indicates performance at the specific point of time of the Web system, and predicts the occurrence of performance overflow which makes it impossible to maintain predetermined performance in the Web system, by comparing the performance prediction value with a predetermined threshold value. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a technique for monitoring the performance of a web system that provides a predetermined service on a network.

Service requests that users transmit to a web system via the Internet using a web browser or the like are increasing dramatically year by year. In recent years, there are frequent cases where the system cannot keep up with the increase in traffic volume and the system goes down. Therefore, there is a need for web system performance monitoring technology that can monitor the performance of web systems and predict the occurrence of performance overflow in advance, as well as quickly detect the occurrence of performance overflow and implement appropriate countermeasures. . As such a technique, for example, Patent Document 1 discloses a WWW site that predicts site performance by periodically predicting fluctuations in the site access of the WWW site and determines whether or not the predetermined performance can be maintained. A performance monitoring apparatus is disclosed.
JP 2002-268922 A

  Web systems are usually composed of a very large number of servers, and the services provided are often diverse. Therefore, even if a single model is created for the entire web system, only the performance of the entire system can be grasped, and the location where the performance overflow has occurred cannot be identified. There is a problem that it is difficult to implement.

  The present invention has been made in view of such circumstances, and an object thereof is to provide a web system performance monitoring apparatus that predicts the performance of a web system using a prediction model.

  One embodiment of the present invention is a web system performance monitoring apparatus. This apparatus stores a model that stores a data acquisition unit that acquires predetermined parameters from a monitored web system that provides a service on a network, and a base model that is a base of a prediction model for predicting the performance of the web system. A factor input unit that accepts input of factors that are data that affects the prediction results using the prediction model, a prediction model creation unit that modifies the base model using the factor and creates a prediction model, The performance prediction value, which is data representing the performance of the web system at a specific point in time, is obtained from the predicted model and parameters, and the web system cannot maintain the predetermined performance by comparing the performance prediction value with a predetermined threshold value. When a performance overflow is predicted, and when a performance overflow is predicted Comprising a countermeasure execution unit for executing a predetermined countermeasure processes to recover the performance of a web system, a.

  According to this aspect, the factors that affect the accuracy of the prediction model for predicting the performance of the web system are specified, and the prediction model reflecting the factors is created. This makes it possible to predict the performance of the web system at a specific point in the future with high accuracy. Therefore, it is possible to take in advance measures to predict and prevent the occurrence of overflow of performance.

  When the web system includes a plurality of application servers that provide different services, and each service is realized by a combination of a plurality of applications, the model storage unit predicts the performance of each of the services, applications, and servers. The base model that is the base of the model is stored, the predictive model creation unit creates a predictive model for each of the service, application, and server, and the full-performance predictor predicts the created service, application, and server. You may obtain | require the performance prediction value matched beforehand with each prediction model using a prediction model.

  By creating a predictive model for each application, service realized by a combination of applications, or for each server, which hardware or software in the web system is a performance bottleneck, and in the web system It is easy to discover which hardware or software can cause performance overflow. Therefore, it is possible to take appropriate avoidance measures against the predicted performance overflow. The same applies when the web system includes a plurality of servers that share services for each access route to the web system.

  It should be noted that any combination of the above-described components and a representation of the present invention by a method, apparatus, system, recording medium, and computer program are also effective as an aspect of the present invention.

  According to the present invention, it is possible to create a prediction model of a web system that provides a service on a network and predict future performance in the web system.

  One embodiment of the present invention is a performance monitoring apparatus that monitors the performance of a web system that provides a service on a network. This performance monitoring apparatus predicts the performance of a future web system by creating a performance prediction model for the web system and inputting parameters obtained from the web system to the prediction model.

  In the following, the present embodiment will be described in detail by taking as an example a performance monitoring device for a securities system that processes securities trading orders and provides information relating to trading.

  FIG. 1 shows a configuration of a securities system 100 that is a monitoring target of the performance monitoring apparatus according to the present embodiment. The securities system 100 is a system that is installed in a securities company and accepts securities trading orders from customers and provides various information useful for securities orders.

  The securities system 100 includes a presentation layer 110 and a business logic layer 120. The presentation layer 110 mainly receives data input from the client terminal and passes it to the business logic layer 120, or processes the data provided from the business logic layer 120 into a web page and provides the client terminal with the data. . The business logic layer 120 has a role in charge of actual processing such as ordering / buying securities and providing information.

  The presentation layer 110 includes a plurality of web servers. In the present embodiment, a group of servers that execute web page processing is prepared for each access channel to the securities system 100. In FIG. 1, it is assumed that the securities system 100 is accessed from a general customer via the web, a call center, and a branch using each client terminal 82a to 82c. Therefore, the presentation layer 110 is in charge of the general web server 102 in charge of processing for access from general customers, the call center web server 104 in charge of processing in access from the call center, and the processing for access from branch offices. And a branch web server 106. Further, a spare web server 108 that is normally inactive may be prepared. The web servers 102 to 108 may be composed of a plurality of servers. In addition to the above functions, each of the web servers 102 to 108 has a dynamic page generation function and a transaction management function for executing a program in response to a request from a client terminal and transmitting the result.

  In the business logic layer 120, a group of servers that execute a predetermined service exclusively, not for each access channel, is prepared. In the present embodiment, the information providing application server 122 that executes a service for providing information necessary for securities trading such as current and past stock prices, trading volumes, and company information to client terminals, and trading orders such as stocks and investment trusts. It is assumed that an order processing application server 124 that executes a service for accepting services, and a business processing application server 126 that executes services for managing user login and logout, customer information, and the like are provided. In addition to this, for example, an external linkage application server that provides economic news in cooperation with an external newspaper company site, a report application server that provides a transaction balance report for each customer, etc. Also good. The application servers 122 to 126 may be composed of a plurality of servers. In addition to the business processing functions described above, each of the application servers 122 to 126 also implements functions that general application servers have, such as a connection function to a database, a transaction management function that links a plurality of processes, and the like. Yes.

  In each of the application servers 122 to 126, services 130 to 140 for a plurality of client terminals are simultaneously executed. For example, in the information providing application server 122, a service A 130 that provides stock price information to general customers and a service B 132 that provides stock price information to a call center may be executed simultaneously. In the order processing application server 124, a service C134 for processing a stock trading order from a call center and a service D136 for processing an investment trust trading order from a store may be executed simultaneously.

  In this specification, a program that realizes each function is called an “application”, and a program that is realized by a combination of applications is called a “service”. For example, the “order processing service” executed by the order processing application server 124 includes applications of “order restriction stock check”, “purchasing power check”, “customer attribute check”, and “order processing”.

  General customers, call center and branch office staff access the securities system 100 via a network 80 such as the Internet using software such as a web browser from the respective client terminals 82a to 82c. The client terminals 82 b and 82 c provided in the call center or branch may be connected to the securities system 100 by a dedicated line 84.

  When the securities system 100 is accessed from the client terminals 82a to 82d, web pages created by the web servers 102 to 106 corresponding to the access channel of the presentation layer 110 are provided to the client terminals 82a to 82d. The client terminals 82a to 82d display the provided web page on the display. When the user performs operations such as login, information browsing, and sales order on the web page, the information is passed to the application servers 122 to 126 of the business logic layer 120 via the web servers 102 to 106. The processing results of the application servers 122 to 126 are processed into web pages by the web servers 102 to 106, and the data is transmitted to the client terminals 82a to 82d. In this way, the user can refer to various information displayed on the web page, or place an order for buying and selling securities through the web page.

  FIG. 2 is a functional block diagram showing the configuration of the performance monitoring apparatus 10. These configurations can be realized in terms of hardware by a CPU, memory, and other LSIs, and in terms of software, they are realized by programs loaded in the memory. Here, functional blocks realized by their cooperation Is drawn. Accordingly, those skilled in the art will understand that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof.

  The data acquisition unit 12 acquires predetermined parameters from the monitored securities system 100. This parameter relates to the performance of the hardware constituting the web system, the operation in the running application, the usage status of the service provided, and the like. The data acquisition unit 12 may change the acquisition timing according to the type of parameter. Specific parameter examples and acquisition timing examples will be described later with reference to FIG.

  The model storage unit 18 stores a base model that is a base of a prediction model for predicting the performance of the securities system 100. A prediction model is created by correcting this base model according to the usage situation of the securities system 100 that changes every moment. In the model storage unit 18, different base models are prepared for servers, applications, and services. The base model is created in advance by the system administrator in consideration of the usage status and operation status of the past securities system 100, hardware configuration, type of application to be executed, performance, future system usage prediction, etc. 18.

  The factor input unit 14 receives an input of a factor for correcting the prediction model. In this specification, the “factor” refers to data that can affect the accuracy of prediction in a prediction model of a securities system. Some factors are manually input by the system administrator, while others are automatically acquired from the securities system 100. An example of the factor will be described later with reference to FIG.

  The prediction model creation unit 16 extracts the base model from the model storage unit 18, modifies the base model using the factor received from the factor input unit 14, and creates a prediction model. The “prediction model” is a model that can predict the performance of the securities system 100 at a specific future time point by inputting one or more parameters at the present time. The base model stored in the model storage unit 18 is common to servers, applications, and services. However, the prediction model creation unit 16 includes a server included in the securities system 100 and an application executed on the securities system 100. It is preferable that a separate prediction model is created for each service. The base model stored in the model storage unit 18 may be a different model for each of the server, application, and service. In the example of FIG. 1, the “resource prediction model” is set for each of the web servers 102 to 106 or the application servers 122 to 126, and the “service prediction model” is set for each of the services 130 to 140 executed by the application servers 122 to 126. ”Is created for each application included in each service.

  When creating a resource prediction model, an application prediction model, and a service prediction model, the prediction model creation unit 16 modifies the base model using different factors. The correspondence between the factor and the prediction model to be corrected is determined in advance.

  The performance overflow prediction unit 22 receives the parameter acquired by the data acquisition unit 12. Further, the performance overflow prediction unit 22 receives the resource prediction model 24, the application prediction model 26, and the service prediction model 28 from the prediction model creation unit 16. And a parameter is input into the created prediction models 24-28, and the performance prediction value which is the data showing the performance in the specific time of the securities system 100 is calculated. The performance prediction value includes, for example, service processing time, the number of requests from the user, turnaround time, calculation load, storage free space, and the like, but is not limited thereto. All or some of the performance prediction values may be the same as the parameters acquired by the data acquisition unit 12.

  Note that parameters to be input to each of the prediction models 24 to 28 are predetermined for each prediction model. Moreover, the performance prediction value calculated by each prediction model 24-28 is also predetermined for every prediction model. For example, from the resource prediction model 24, the calculation load and the storage free capacity can be obtained from the above performance prediction values.

  Furthermore, the performance overflow prediction unit 22 can maintain the desired performance of the securities system 100 at a specific time in the future by comparing the calculated performance prediction value with a threshold value determined for each performance prediction value. It is determined whether or not. In this specification, a state in which the securities system 100 cannot maintain desired performance is referred to as “performance overflow”. Examples of performance overflows include increased service processing time and turnaround time due to increased traffic, the number of requests from users exceeding the threshold, and the computational load exceeding the threshold. However, the storage free capacity may be less than the threshold value, but is not limited thereto.

  The countermeasure execution unit 30 executes a predetermined countermeasure process for recovering the performance overflow of the securities system 100 when the performance overflow prediction unit 22 determines that the performance overflow will occur at a specific time in the future. The countermeasure execution unit 30 includes a resource allocation unit 32, a login restriction unit 34, a service stop unit 36, and an error notification unit 38.

  When it is determined that the performance overflow occurs due to a shortage of resources for executing the application or service in the securities system 100, the resource allocation unit 32 predicts that the server prepared in advance will run out of resources. Or distribute to services. To describe a more specific example, for example, in the performance overflow prediction unit 22, in the resource prediction model of any of the web servers 102 to 106, the calculation load that is the performance prediction value exceeds the threshold value, or the storage free capacity Is predicted to fall below the threshold. In this case, the resource allocation unit 32 instructs the spare web server 108 to share part of the processing in the web server that is predicted to overflow.

  Alternatively, the resource allocation unit 32 may distribute unused servers or workspaces in the securities system 100 to applications or services that are predicted to have insufficient resources. In this case, the data acquisition unit 12 needs to detect unused resources in the securities system 100 in advance. As described above, the resources in the system can be dynamically operated by the resource allocation unit 32.

  The login restriction unit 34 restricts login to the securities system 100 when it is determined that the performance overflow occurs due to a shortage of resources for executing applications or services in the securities system 100. Specifically, the login restriction unit 34 instructs the business processing application server 126 not to accept a certain number of logins.

  The service stopping unit 36 stops providing an application or service other than the specific application or service when it is determined that performance overflow occurs due to a shortage of resources for executing the application or service in the securities system 100. This particular application or service is preferably one that performs the most important core business in the system. In the case of the securities system 100, the buying and selling order service in the order processing application server 124 is most important. Therefore, a service for providing information in the information providing application server 122 or a service for updating customer information by the business processing application server 126 is provided. It is preferable to stop. By stopping some services in this way, it is possible to prevent the entire system from being down due to overflowing performance while maintaining the core service.

  The service stopping unit 36 may stop only the web server corresponding to a specific access channel when it is determined that the securities system 100 has insufficient resources. For example, the general web server 102 may be stopped and the call center web server 104 and the branch web server 106 may be maintained as they are. In this case, the resource allocation unit 32 may be instructed to allocate the resources of the general web server 102 to the call center web server 104 or the branch web server 106.

  When it is determined by the performance overflow prediction unit 22 that the security system 100 will overflow, the error notification unit 38 sends an e-mail to the effect that the performance overflow is predicted to the system administrator or the like. As a result, the system administrator can quickly take measures before the performance overflow actually occurs. The error notification unit 38 may instruct the web servers 102 to 106 to display a request for the user such as “please refrain from connecting for a while” on the web page.

  The countermeasure execution unit 30 is a function for formulating a recovery procedure when an actual performance overflow occurs in the securities system 100 and a system augmentation procedure when a future traffic increase is expected and presenting it to the system administrator. You may have.

The system diagnosis unit 40 periodically inspects the performance of the securities system 100 to diagnose whether or not a desired performance is obtained. For example, the processing speed or turnaround time for each application is measured and compared with a predetermined threshold value or the previous measurement result. If the desired performance is not achieved, system improvement measures such as review of application and database design are formulated and presented to the system staff.
The system diagnosis unit 40 compares the actual performance in the securities system 100 with the performance calculated using the prediction model in the performance overflow prediction unit 22, and reduces the error between the actual value and the predicted value. It may have a function of correcting the prediction model.

  The performance overflow detection unit 20 monitors the operation of the securities system 100 in real time. For example, the processing load and traffic volume of each server are monitored, and a warning is issued to the system administrator when a predetermined threshold value is exceeded. When a performance overflow is detected by the performance overflow detection unit 20, the countermeasure execution unit 30 executes a countermeasure process similar to that when a performance overflow is predicted.

Next, a prediction model creation process in the prediction model creation unit 16 will be described with reference to the flowchart of FIG.
First, the prediction model creation unit 16 reads a base model prepared for each server, application, and service from the model storage unit 18 (S10). Subsequently, the factor input by the factor input unit 14 is received (S12). Furthermore, the prediction model creation unit 16 acquires hardware performance values of each server in the securities system 100 in units of machines and groups (S14). This hardware performance value is a target value of performance to be exhibited by the server and the server group. The hardware performance value may be input in advance by the system administrator, or may be automatically measured in a time zone when the processing load of the securities system 100 is low. The prediction model is created on the assumption that each server operates based on the hardware performance value acquired by each server. The reason why hardware performance values are acquired in units of machines and groups is that the performance exhibited when a plurality of servers operate in cooperation often does not simply become a total.

  The prediction model creation unit 16 modifies the base model using the acquired factor and hardware performance value, and creates a resource prediction model, an application prediction model, and a service prediction model (S16). As a method of correcting the base model based on the factor and the hardware performance value, for example, a change of a proportional coefficient in a mathematical expression constituting the base model, a deletion or addition of a specific term, and the like can be considered.

  Finally, the prediction model creation unit 16 sets a threshold value for determining that the performance overflows for each prediction model (S18). This threshold value may be predetermined for each prediction model, or may change in proportion to a factor. The created prediction model and threshold value are passed to the performance overflow prediction unit 22.

  As described above, by creating a prediction model for each server, each application, and each service, it becomes possible to perform performance prediction and performance overflow countermeasures for each application or service.

  FIG. 4 is a table showing an example of factors considered when creating a prediction model. The table 50 includes a column 52 that represents the type of factor, a column 54 that represents the factor, and a column 56 that represents a prediction model to be corrected in consideration of the factor. Predictive models vary from those that are heavily influenced by certain factors to those that are largely unaffected. Therefore, it is preferable to determine a prediction model to be corrected for each factor.

  Factor types include stationary factors, periodic factors, and sudden factors. The stationary factor is a factor that needs to be acquired at regular intervals during the operation of the securities system 100. The regular factors include the number of contract accounts, the number of orders, the trading volume, the number of logins per access channel, and the number of requests. Among these factors, the factor of the “contract account number” greatly affects the performance of the paperwork application related to the management of user information, so the prediction model of the paperwork application is to be corrected. All other factors greatly affect the order processing volume, and the order processing application is subject to correction.

  The period factor is a factor that needs to be considered only for a certain period (for example, one month). Periodic factors include the provision of new services, the provision of high-speed lines, and the implementation of sales. All of these are considered to lead to an increase in the traffic volume for a certain period after provision, and thus correction such as increasing the proportional coefficient of the prediction model is necessary only during that period. Due to its nature, the period factor must be entered manually by the system administrator.

  An abrupt factor is a factor that needs to be considered only momentarily (eg, for one day). Sudden factors include the occurrence of specific incidents and the recovery of system failures. Since these are considered to lead to an increase in traffic momentarily, it is necessary to make corrections such as increasing the proportionality coefficient of the prediction model only during that time. Sudden factors must also be entered manually by the system administrator.

  As described above, by associating a factor with a prediction model to be corrected in consideration of the factor, the prediction model can be appropriately corrected.

FIG. 5 is a table showing an example of parameters input to the prediction model. The table 60 includes a column 62 representing a prediction model to which parameters are inputted, a column 64 representing parameters, and a column 66 representing parameter acquisition timing.
The parameters input to the service prediction model include the number of requests and the number of logins. These acquisition timings are, for example, every minute.
Parameters input to the application prediction model include the number of application calls, processing time, turnaround time, page view, and the like. The acquisition timing of the number of calls is a predetermined time, and the acquisition timing of other parameters is, for example, every hour.
Parameters input to the resource prediction model include a CPU usage rate, a memory usage amount, a storage free space, and the like. These acquisition timings are, for example, every minute.

FIG. 6 is a flowchart of the performance overflow prediction process in the performance overflow prediction unit 22.
First, the performance prediction unit 22 receives a resource prediction model, an application prediction model, a service prediction model, and respective threshold values from the prediction model creation unit 16 (S30). Subsequently, a parameter corresponding to each prediction model is input, and a predicted performance value at a specific time in the future is calculated (S32). The performance overflow prediction unit 22 compares the calculated performance predicted value with a threshold value (S34), and determines whether or not an occurrence of performance overflow is predicted in the securities system 100 (S36). If the occurrence of performance overflow is not predicted (N in S36), the process is terminated. If the occurrence of performance overflow is predicted (Y in S36), the countermeasure execution unit 30 executes the performance overflow recovery process (S38). ).

  FIG. 7 shows the hardware configuration of the web server and application server of the securities system 100. Each server includes a CPU 70 that executes various processes according to a program, a memory 72 that temporarily stores data and programs, a storage device 74 such as a hard disk drive and a DVD disk drive, and various input / output processes connected to a network. At least a network interface 76 for performing the above and a bus 78 for interconnecting them. The server may have an input device such as a keyboard and a mouse and an output device such as a display as necessary. One server may have two or more network interfaces 76.

  Each server is preferably a blade server in which elements necessary for operating as a computer on a single board, that is, a CPU, a memory, a hard disk, a bus, and the like are mounted, and the web servers 102 to 108 and application servers 122 to 126 preferably takes a configuration in which a plurality of blade servers are inserted into the housing, but may take other forms.

  Each server is configured to be able to communicate with a network storage (not shown) via a router (not shown). A master table, a database, and the like necessary for executing the application are stored in the network storage, and can be transmitted from the network storage to each server as necessary. The network storage is generally a hard disk device, and normally a RAID is formed by collecting a plurality of disks, but may be a magnetic tape device.

  As described above, according to the present embodiment, a prediction model for predicting performance at a specific time in the future is created for each of a plurality of servers, applications, and services. It is possible to easily discover which hardware or software in the system can cause the overflow of system performance. Therefore, it is possible to take appropriate avoidance measures according to the predicted performance overflow.

  In the conventional web system monitoring device, it is difficult to predict the impact on the system due to an increase in traffic, and therefore, measures such as system enhancement cannot be implemented quickly. In contrast, in the performance monitoring apparatus of the present embodiment, factors that affect the accuracy of the prediction model are specified, and these factors are reflected in the creation of the prediction model. This makes it possible to predict the performance of the securities system at a specific point in the future with high accuracy. Therefore, instead of responding after the occurrence of overflow, it is possible to predict the occurrence of overflow and take measures to prevent it.

  In addition, a base model is prepared for each server, application, and service, and the base model is modified by using factors that match the characteristics of each to create a prediction model. This improves the accuracy of system performance prediction. To do.

  In addition, since the factor types are divided into a plurality, and the steady factor is constantly acquired and reflected in the model, a change in the system usage status can be quickly reflected in the prediction model. Furthermore, since a period factor and an abrupt factor are provided and a factor that needs to be considered only for a limited period is reflected in the prediction model, the accuracy of system performance prediction is improved.

  The present invention has been described based on some embodiments. Those skilled in the art will understand that these embodiments are exemplifications, and that there may be various modifications to the combinations of the respective constituent elements and processing processes, and such modifications are also within the scope of the present invention. By the way.

  It should also be understood by those skilled in the art that the functions to be fulfilled by the constituent elements described in the claims are realized by the individual functional blocks shown in the present embodiment or their linkage.

  In the embodiment, it is described that a prediction model is created for a service. However, a prediction model may be created for a “service group” that is a combination of a plurality of services. Here, the service group includes a series of services for completing the business flow. For example, a service group including all of “stock price code input service”, “market price confirmation service”, “purchase unit / method input service”, and “order reception service” can be considered.

  In the embodiment, the securities system is exemplified as the monitoring target of the performance monitoring device, but the web system to be monitored is not limited thereto.

It is a figure which shows the structure of the securities system which is the monitoring object of the performance monitoring apparatus concerning one Embodiment of this invention. It is a functional block diagram of the performance monitoring device of one embodiment. It is a flowchart which shows the preparation process of the prediction model in a model preparation part. It is a figure which shows an example of the factor considered in the case of preparation of a prediction model. It is a figure which shows an example of the parameter input into a prediction model. It is a flowchart of the performance overflow prediction process in a performance overflow prediction part. It is a figure which shows the hardware constitutions of each server.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 10 Performance monitoring device, 12 Data acquisition part, 14 Factor input part, 16 Prediction model creation part, 18 Model storage part, 20 Performance overflow detection part, 22 Performance overflow prediction part, 24 Resource prediction model, 26 Application prediction model, 28 Service Prediction model, 30 countermeasure execution unit, 32 resource allocation unit, 34 login restriction unit, 36 service stop unit, 38 error notification unit, 80 network, 82a-d client terminal, 100 securities system, 102 general web server, 104 call center Web server, 106 Branch web server, 122 Information providing application server, 124 Order processing application server, 126 Business processing application server, 130-140 -Bis.

Claims (10)

A data acquisition unit that acquires predetermined parameters from a monitored web system that provides a service on a network;
A model storage unit that stores a base model serving as a base of a prediction model for predicting the performance of the web system;
A factor input unit that accepts an input of a factor that is data that affects the prediction result using the prediction model;
A prediction model creation unit that creates the prediction model by correcting the base model using the factor;
A performance prediction value, which is data representing the performance of the web system at a specific point in time, is obtained from the created prediction model and the parameter, and a predetermined value is determined in the web system by comparing the performance prediction value with a predetermined threshold value. A performance overflow prediction unit that predicts the occurrence of performance overflow that can no longer maintain the performance,
When a performance overflow is predicted, a countermeasure execution unit that executes a predetermined countermeasure process for recovering the performance of the web system;
A web system performance monitoring apparatus comprising: When the web system includes a plurality of application servers that provide different services, and each service is realized by a combination of a plurality of applications,
The model storage unit stores a base model that is a base of a prediction model for predicting the performance of each service, application, and server,
The prediction model creation unit creates a prediction model for each of a service, an application, and a server,
The said performance overflow prediction part calculates | requires the performance prediction value matched beforehand with each prediction model using the prediction model of the created service, an application, and a server. Performance monitoring device. When the web system includes a plurality of servers that share services for each access route to the web system,
The model storage unit stores a base model that is a base of a prediction model for predicting the performance of each of the plurality of servers,
The prediction model creation unit creates a prediction model for each of the plurality of servers,
The performance monitoring apparatus according to claim 1, wherein the performance overflow prediction unit obtains a performance prediction value associated with each prediction model in advance using the created prediction model.   4. The performance monitoring apparatus according to claim 1, wherein the prediction model creation unit corrects a different prediction model according to a type of a factor input in the factor input unit. 5.   The performance monitoring apparatus according to claim 1, wherein the data acquisition unit changes an acquisition timing according to a type of the parameter.   The said performance overflow prediction part uses the said parameter and the said prediction model by which correspondence is decided beforehand according to the performance prediction value to obtain | require, The Claim 1 thru | or 3 characterized by the above-mentioned. Performance monitoring device. When the performance overflow prediction unit predicts that a performance overflow occurs due to a shortage of resources for executing applications or services in the web system,
The said countermeasure execution part is provided with the resource allocation part which distributes the server prepared beforehand beforehand to the application or service estimated that the resource is insufficient in the said web system. The performance monitoring device according to the above. When the performance overflow prediction unit predicts that a performance overflow occurs due to a shortage of resources for executing applications or services in the web system,
The data acquisition unit detects unused resources in the web system,
The said countermeasure execution part is provided with the resource allocation part which distributes the detected unused resource to the application or service with which it was estimated that the resource runs short in the said web system. The performance monitoring device according to the above. When the performance overflow prediction unit predicts that a performance overflow occurs due to a shortage of resources for executing applications or services in the web system,
The performance monitoring apparatus according to claim 1, wherein the countermeasure execution unit includes a login restriction unit that restricts login to the web system. When the performance overflow prediction unit predicts that a performance overflow occurs due to a shortage of resources for executing applications or services in the web system,
The said countermeasure execution part contains the service stop part which stops provision of another application or service, in order to maintain provision of a part of application or service in the said web system, The 1 thru | or 6 characterized by the above-mentioned. The performance monitoring device according to any one of the above.

Images