CN105868004B - Scheduling method and scheduling device of service system based on cloud computing - Google Patents

Abstract

The present invention provides a scheduling method and a scheduling device for a cloud computing-based business system. By analyzing the historical load, it can predict the future business situation, so as to conduct virtual machine scheduling scientifically and efficiently. The invention can simplify the scheduling algorithm of the multi-service cloud system, schedule the multi-service cloud system more directly and efficiently, reduce the complexity of the scheduling algorithm, and reduce the scheduling "jitter" problem caused by the traditional scheduling algorithm, so as to avoid the problem of "jitter" in the business peak. The performance degradation of the business system caused by the necessary virtual machine migration.

Description

A scheduling method and scheduling device for a business system based on cloud computing

technical field

The invention relates to a cloud computing resource scheduling method, in particular to a scheduling method and a scheduling device of a cloud computing-based business system.

Background technique

Telecom operators' value-added service systems, such as SMS center, MMS center, WAP gateway and other service systems, are an important source of operators' income, so the construction of value-added service systems has always been paid attention by operators. In the traditional construction scheme, each set of value-added business systems uses dedicated server resources. For example, the server used by the MMS center is exclusively owned by the MMS center. The server used by the SMS center is exclusively owned by the SMS center. An obvious disadvantage of this independent chimney construction method is the serious waste of physical server resources.

With the development of virtualization and cloud computing technologies, the independent, chimney-like construction mode of value-added service systems has been gradually replaced by cloud computing. In the cloud computing mode, resource pooling is first performed, that is, through virtualization technology, the physical server is virtualized into multiple virtual machines, and the business system is carried on the virtual machines. After the physical server resources are pooled in this way, a virtualization layer is added to decouple the coupling relationship between the business system and the physical server, making it possible for multiple business systems to reuse a set of physical servers. This multi-service cloud system built with cloud computing technology not only improves the utilization of resources, but also provides the possibility to save energy, reduce consumption, and reduce operation and maintenance costs.

The service system of telecom operators has obvious tidal characteristics. For example, during the day, the load of the service system shows a peak due to the strong communication demand of people, and the load of the service system will drop significantly at night due to the decrease of people's communication demand. Therefore, multi-service cloud systems using cloud computing technology often use the migration characteristics of virtualization to perform intelligent virtual machine scheduling to reduce power consumption and cost. A typical solution in the prior art is that, during the day, all virtual machines carrying services are evenly distributed among each server in the resource pool, and the system carries the arrival of the peak service in the best performance state. When the load is low at night, the virtual machines that carry services are compressed and migrated to the fewest servers to run. At the same time, the idle physical servers that do not run virtual machines are powered off or hibernated to reduce the power consumption of these servers. achieve the purpose of reducing costs.

However, the problem with the current scheduling algorithm is that it lacks global consideration, only considers the resource load at the scheduling moment, and does not take into account the difference between different business loads in the business system, which will lead to scheduling again during the middle operation of the system. A scheduling jitter problem arises.

SUMMARY OF THE INVENTION

The technical problem to be solved by the embodiments of the present invention is to provide a scheduling method and a scheduling device for a business system based on cloud computing, so as to reduce the complexity of the scheduling algorithm and reduce the scheduling jitter problem caused by the traditional scheduling algorithm.

In order to solve the above technical problems, the scheduling method of the cloud computing-based business system provided by the embodiment of the present invention includes:

When the business system is switched to a scene mode, a pre-saved mapping relationship table between the virtual machine of the business system and the physical server in the scene mode is obtained, and the business system is recorded in the mapping relationship table The mapping relationship between the virtual machine and the physical server obtained by scheduling according to the predetermined virtual machine scheduling algorithm in the scenario mode last time, and the scenario mode corresponds to a periodic time period of the business system;

According to the mapping relationship in the mapping relationship table, the virtual machine of the business system is scheduled to the corresponding physical server.

Among them, in the above method,

When the business system is switched to the scene mode, if the mapping table in the scene mode has not been saved, the resource occupancy of each physical server in the business system is periodically collected, and the resources are collected according to a predetermined schedule. The virtual machine scheduling algorithm is used to execute the migration scheduling of virtual machines;

And, according to the mapping relationship between the virtual machine and the physical server after migration and scheduling, the mapping relationship table in the scenario mode is recorded and saved.

Wherein, in the above method, the scheduling of the virtual machine of the business system to the corresponding physical server according to the mapping relationship in the mapping relationship table includes:

Wake up the physical server recorded in the mapping relationship table, and schedule the virtual machine recorded in the mapping relationship table to the corresponding physical server according to the mapping relationship in the mapping relationship table; and,

Power off or hibernate the remaining physical servers in the business system except the physical servers recorded in the mapping relationship table.

Wherein, in the above method, after the step of scheduling the virtual machine of the business system to the corresponding physical server according to the mapping relationship in the mapping relationship table, the method further includes:

Periodically collect the resource occupancy status of each physical server in the business system, and perform migration scheduling of virtual machines according to the resource occupancy status and a predetermined virtual machine scheduling algorithm;

And, after executing the migration scheduling of the virtual machine, the mapping relationship table is updated according to the current mapping relationship between the virtual machine and the physical server in the business system.

Among them, in the above method,

The working time period of the business system includes a preset and periodically presented busy time period and an idle time period;

The performing migration scheduling of the virtual machine according to the resource occupancy situation and according to a predetermined virtual machine scheduling algorithm, including:

When the scenario mode is a busy-hour scenario corresponding to the busy-hour period, according to a load balancing algorithm, some virtual machines on the physical server whose resource occupancy rate exceeds a predetermined threshold are migrated to the resource occupancy in the business system on the physical server with the lowest rate;

When the scenario mode is an idle scenario corresponding to the idle time period, the virtual machine of the business system is migrated to a minimum number of physical servers according to a backpack algorithm.

The embodiment of the present invention also provides a scheduling device for a cloud computing-based business system, including:

an acquiring unit, configured to acquire a pre-saved mapping relationship table between the virtual machine of the business system and the physical server in the scenario mode when the business system is switched to a scene mode, and the mapping relationship table records There is a mapping relationship between the virtual machine and the physical server that the business system dispatched according to the predetermined virtual machine scheduling algorithm in the scenario mode last time, and the scenario mode corresponds to a periodic time period of the business system. ;

The first scheduling unit is configured to schedule the virtual machine of the business system to the corresponding physical server according to the mapping relationship in the mapping relationship table.

Wherein, in the above scheduling device, it also includes:

The second scheduling unit is configured to periodically collect the data of each physical server in the business system if the mapping table in the scene mode has not been saved when the business system is switched to the scene mode. Resource occupancy, and perform virtual machine migration scheduling according to the predetermined virtual machine scheduling algorithm;

The first updating unit is configured to record and save the mapping relationship table in the scenario mode according to the mapping relationship between the virtual machine and the physical server after migration and scheduling.

Wherein, in the above scheduling device, the first scheduling unit includes:

a migration unit, configured to wake up the physical server recorded in the mapping relationship table, and schedule the virtual machine recorded in the mapping relationship table to the corresponding physical server according to the mapping relationship in the mapping relationship table; and,

The energy saving unit is used for powering off or hibernating the remaining physical servers in the service system except the physical servers recorded in the mapping relationship table.

Wherein, in the above scheduling device, it also includes:

a third scheduling unit, configured to periodically collect the resource occupancy status of each physical server in the business system, and perform migration scheduling of virtual machines according to the resource occupancy status and a predetermined virtual machine scheduling algorithm; and, in After the migration scheduling of the virtual machine is executed, the mapping relationship table is updated according to the current mapping relationship between the virtual machine and the physical server in the business system.

Wherein, in the above scheduling device, the working time period of the business system includes a preset and periodically presented busy time period and an idle time period;

The second scheduling unit or the third scheduling unit is further configured to, when the scene mode is a busy-hour scene corresponding to the busy-hour period, according to a load balancing algorithm, the physical server whose resource occupancy rate exceeds a predetermined threshold Some virtual machines on the server are migrated to the physical server with the lowest resource occupancy rate in the business system; and, when the scene mode is an idle time scene corresponding to the idle time period, according to the backpack algorithm, all The virtual machines of the above business systems are migrated to the minimum number of physical servers.

Compared with the prior art, the scheduling method and scheduling device of the cloud computing-based business system provided by the embodiments of the present invention can simplify the scheduling algorithm of the business cloud system, can schedule the business system more directly and efficiently, and reduce the number of scheduling algorithms. It also reduces the scheduling "jitter" problem caused by traditional scheduling algorithms, reduces or avoids the performance degradation of the business system caused by unnecessary virtual machine migration during business peaks, and improves the stability of the business system.

Description of drawings

1 is a schematic flowchart of a scheduling method based on a business system provided by an embodiment of the present invention;

FIG. 2 is a schematic structural diagram of a scheduling apparatus based on a service system provided by an embodiment of the present invention.

Detailed ways

In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

The multi-service cloud system built with cloud computing technology not only improves the utilization rate of resources, but also provides the possibility to save energy, reduce consumption, and reduce operation and maintenance costs. A telecom operator's service system has obvious tidal characteristics. A scheduling scheme adopted for the above tidal characteristics is as follows, that is, during the day, as the load increases, all servers are powered on. Then, all virtual machines that carry services are migrated to each server in the system on average using the scheduling principle of load balancing. At night, an energy-efficient scheduling algorithm scheme, typically a knapsack algorithm, is used to compress the virtual machines to run on the fewest servers.

The above scheduling scheme lacks global consideration, only considers the resource load at the time of scheduling, and does not consider the difference in the load of the business system, which will bring about scheduling jitter again during the intermediate operation of the system. For example, suppose that when the business system spreads virtual machines to all servers according to the principle of CPU balancing, the CPU load considered is only the CPU load at the scheduling moment. Over time, the load peak of each business system will gradually come. Since the load of each business system is different, for example, the virtual machine hosting the MMS center has a peak CPU utilization rate of 45%; the virtual machine hosting the SMS center has a peak CPU utilization rate of 55%, then if When the virtual machines of the two SMS centers are scheduled to run on one physical server, after the peak of the service load arrives, it is likely to trigger rescheduling, and one of the virtual machines running in them is scheduled to go away, such as migrating the MMS virtual machine to another physical server. server, that is, scheduling jitter occurs. This scheduling jitter phenomenon is unavoidable in the scheduling algorithm of the prior art.

The scheduling of traditional operators' multi-service cloud systems does not take into account the differences in the load of each service system, nor does it predict the load of the service system. Therefore, it is easy to cause the virtual machine that carries the service system during the peak of the service. The physical server is overloaded. If the value is high, the scheduled migration is performed by the cloud management system. The virtual machine migration that occurs during the peak of the business will cause the performance of the entire business system to decline due to the performance degradation of the virtual machine itself during the migration, which will lead to the instability of the entire business system.

In order to solve the above problem, the embodiment of the present invention proposes a new scheduling algorithm to reduce or avoid the above-mentioned scheduling jitter phenomenon. machine scheduling. By using the scheduling method described in the embodiment of the present invention, the scheduling algorithm of the multi-service cloud system can be simplified, the multi-service cloud system can be scheduled more directly and efficiently, the complexity of the scheduling algorithm is reduced, and the scheduling "jitter" caused by the traditional scheduling algorithm is reduced. ” problem to avoid the performance degradation of the business system caused by unnecessary virtual machine migration during business peaks.

After researching the existing business system, it is found that when the business system is running, if the load balancing scheduling algorithm is used for scheduling during the day, a steady state will eventually be reached, that is, the virtual machine will no longer be affected by the load change of the business system. The migration of the virtual machine is called steady state at this time. The relationship between all physical servers at this time and all virtual machines running on them is mapped according to the affiliation. A steady-state mapping table is formed, corresponding to the daytime pattern.

Similarly, at night, energy-saving algorithms including the knapsack algorithm are used to compress the virtual machine to run on as few physical servers as possible. After the compression is completed and a steady state is reached, the virtual machine at this time is also mapped to the physical server it is running on, and the mapping relationship table at this time is recorded, corresponding to the night mode.

When the business system is switched from the night mode to the day mode, the scheduler directly migrates the virtual machine to the physical server to which it belongs according to the previously recorded mapping table between the virtual machine and the physical server in the day mode. Such scheduling is equivalent to restoring the system to the last steady state. Since the loads of various existing service systems, including telecommunication service systems, are continuous for a period of time and have specific periodicity and regularity, the steady state of yesterday can usually be predicted, and the same is true for today's Steady-state, or maximally-probable steady-state, is the most basic principle of the scheduling method in the embodiment of the present invention.

Similarly, when switching from the day mode to the night mode, the migration scheduling of the virtual machine may also be performed according to the previously recorded mapping relationship table in the steady state in the night mode.

The scheduling method proposed by the embodiment of the present invention predicts the load and virtual machine distribution in the next cycle according to the load and virtual machine distribution in the previous cycle, and can directly migrate the virtual machine to a Steady state, its scheduling algorithm is extremely simple, and the scheduling effect is accurate and efficient. Referring to FIG. 1 , the scheduling method for a cloud computing-based business system provided by an embodiment of the present invention can be applied to a multi-service business system, where the business system includes multiple physical servers and multiple virtual machines, and the virtual machines For processing different services in the service system, the scheduling method includes the following steps:

Step 11: When the business system is switched to a scene mode, obtain a pre-saved mapping relationship table between the virtual machine of the business system and the physical server in the scene mode, and the mapping relationship table records the The mapping relationship between the virtual machine and the physical server obtained by the business system most recently scheduled according to a predetermined virtual machine scheduling algorithm in the scenario mode, the scenario mode corresponds to a periodic time period of the business system.

Here, the working time period of the business system includes a preset and periodically presented busy time period and an idle time period. The scene corresponding to the busy time period is a busy time scene, and the scene corresponding to the free time period is an idle time scene. Specifically, the scene mode in the above step 11 may be the day mode mentioned above, that is, the busy time scene mode corresponding to the business system in the busy time period, or it may be the night mode, that is, the business system corresponds to the idle time period. idle scene mode.

Step 12, according to the mapping relationship in the mapping relationship table, schedule the virtual machine of the business system to the corresponding physical server.

Here, in the above-mentioned step 12, performing the scheduling process according to the mapping relationship in the mapping relationship table may specifically include: waking up the physical server recorded in the mapping relationship table, and if the physical server is already in the wake-up state, it can be directly executed Next, if the physical server is in a power-off or dormant state, wake up the physical server; then, according to the mapping relationship in the mapping relationship table, schedule the virtual machine recorded in the mapping relationship table to the corresponding physical server ; After the scheduling is completed, power off or hibernate the remaining physical servers in the business system except the physical servers recorded in the mapping relationship table to save energy consumption.

Through the above steps, in this embodiment, when the business system switches from one scene mode to another scene mode, the virtual machine of the business system and the physical server are quickly connected according to the most recent pre-recorded mapping relationship table in the other scene mode. The mapping relationship of is restored to the last situation in this scene mode, that is, a steady state is quickly reached, so that the processing of the scheduling algorithm can be simplified, and the scheduling jitter steady state can be reduced or avoided.

In the above step 11, if the saved mapping table in the scene mode cannot be found when the business system switches to the scene mode, it indicates that the scene mode may be entered for the first time, and at this time, it can be based on The affiliation between the current virtual machine and the physical server in the business system, periodically collects the resource occupancy of each physical server in the business system, and executes the migration scheduling of the virtual machine according to the predetermined virtual machine scheduling algorithm to make it reach the a steady state. And, after each migration scheduling occurs, according to the mapping relationship between the virtual machine and the physical server after migration scheduling, record and save the mapping relationship table in the scenario mode, if the mapping relationship table already exists, Then, the mapping relationship table is updated according to the mapping relationship between the virtual machine and the physical server after migration and scheduling.

Since the load of the business system is not constant, after returning to the steady state in the previous scenario mode, it may be necessary to continue to schedule and migrate the virtual machine according to the current change of the business load to adapt to the new changes. At this time, after the above step 12, the embodiment of the present invention may further include the following steps: periodically collecting the resource occupancy status of each physical server in the service system, and according to the resource occupancy status, according to a predetermined virtual machine scheduling algorithm to execute the migration scheduling of the virtual machine; then, after executing the migration scheduling of the virtual machine, the mapping relationship table is updated according to the current mapping relationship between the virtual machine and the physical server in the business system.

In the embodiment of the present invention, according to the resource occupancy situation, according to the predetermined virtual machine scheduling algorithm, when the migration scheduling of the virtual machine is performed, different processing may be performed for different scenarios, for example:

When the scenario mode is a busy-hour scenario corresponding to the busy-hour period, some virtual machines on the physical server whose resource occupancy rate exceeds a predetermined threshold may be migrated to resources in the business system according to a load balancing algorithm On the physical server with the lowest occupancy;

When the scenario mode is an idle scenario corresponding to the idle time period, the virtual machine of the business system may be migrated to a minimum number of physical servers according to a backpack algorithm.

Through the above method, the embodiment of the present invention can simplify the migration scheduling of the virtual machine of the service system, reduce or avoid the steady state of scheduling jitter, and improve the performance and stability of the service system.

Based on the above method, an embodiment of the present invention further provides an apparatus for implementing the above method. Please refer to FIG. 2 , the scheduling device of the cloud computing-based business system provided by the embodiment of the present invention includes:

The acquiring unit 21 is configured to acquire a pre-saved mapping relationship table between the virtual machine of the business system and the physical server in the scenario mode when the business system is switched to a scene mode, in the mapping relationship table The mapping relationship between the virtual machine and the physical server that was last scheduled by the business system in the scenario mode according to a predetermined virtual machine scheduling algorithm is recorded, and the scenario mode corresponds to a periodic time of the business system. part;

The first scheduling unit 22 is configured to schedule the virtual machine of the business system to a corresponding physical server according to the mapping relationship in the mapping relationship table.

Further, the above-mentioned scheduling device may also include:

The second scheduling unit is configured to periodically collect the data of each physical server in the business system if the mapping table in the scene mode has not been saved when the business system is switched to the scene mode. Resource occupancy, and perform virtual machine migration scheduling according to the predetermined virtual machine scheduling algorithm;

The first updating unit is configured to record and save the mapping relationship table in the scenario mode according to the mapping relationship between the virtual machine and the physical server after migration and scheduling.

Specifically, the above-mentioned first scheduling unit 22 may include:

a migration unit, configured to wake up the physical server recorded in the mapping relationship table, and schedule the virtual machine recorded in the mapping relationship table to the corresponding physical server according to the mapping relationship in the mapping relationship table; and,

The energy saving unit is used for powering off or hibernating the remaining physical servers in the service system except the physical servers recorded in the mapping relationship table.

In order to adapt to changes in the service load of the service system, the above-mentioned scheduling device may further include:

a third scheduling unit, configured to periodically collect the resource occupancy status of each physical server in the business system, and perform migration scheduling of virtual machines according to the resource occupancy status and a predetermined virtual machine scheduling algorithm; and, in After the migration scheduling of the virtual machine is executed, the mapping relationship table is updated according to the current mapping relationship between the virtual machine and the physical server in the business system.

Specifically, the working time period of the service system may include a preset and periodically presented busy time period and an idle time period. At this time, the second scheduling unit or the third scheduling unit is further configured to, when the scene mode is a busy-hour scene corresponding to the busy-hour period, according to a load balancing algorithm, make the resource occupancy rate exceed a predetermined threshold Part of the virtual machines on the physical server, migrate to the physical server with the lowest resource occupancy rate in the business system; and, when the scene mode is an idle time scene corresponding to the idle time period, according to the backpack algorithm , migrating the virtual machine of the business system to a minimum number of physical servers.

Hereinafter, the present invention will be described in more detail through a more specific example of a scheduling method. This scheduling method is divided into four stages, namely, the initial day stage, the initial night stage, the normal day stage and the normal night stage.

Initial daytime phase:

Step 1: As the initial state, the business system randomly distributes virtual machines (VMs) to each physical server host (Host) according to the principle of load balancing scheduling algorithm (an optional algorithm is the random spreading algorithm), and establishes VM and Host. The mapping table for the daytime phases.

Step 2: Perform dynamic scheduling according to the load situation. The strategy of telecom operators is generally that if the CPU usage of the Host exceeds 70%, VM scheduling needs to be performed. Therefore, the cloud management scheduling device detects the CPU usage of each host in real time. If the CPU usage exceeds 70% of the host, the VM with the highest CPU usage is selected and migrated to the host with the lightest load. At the same time, update the mapping relationship table between the VM and the Host during the daytime.

Repeat step 2 until the day mode is switched to the night mode.

Initial night phase:

Step 1: As the initial state, the business system compresses the VM (virtual machine) to the minimum Host (physical host) according to the principle of the energy-saving scheduling algorithm (an optional algorithm is the backpack algorithm), and establishes the night phase mapping between the VM and the Host. Relational tables.

normal daytime

Step 1: When switching from night mode to day mode, quickly schedule the VM to the corresponding Host according to the previously established mapping table between the VM and the Host during the daytime phase.

Step 2: Detect the CPU usage of each Host in real time. If the CPU usage exceeds 70% of the Host, select the VM with the highest CPU usage on it, migrate it to the Host with the lightest load, and update the VM and Host during the day. mapping table.

The above step 2 is performed periodically until the day and night modes are switched and the scheduling policy is switched.

Normal night phase:

Step 1: Quickly schedule VMs to corresponding Hosts according to the previously established mapping table between VMs and Hosts at night, and perform hibernation or power-off operations on idle Hosts to save power and air conditioning cooling consumption.

To sum up, the scheduling algorithm and the scheduling device provided by the embodiments of the present invention can effectively schedule the operator's multi-service cloud system, and on the basis of ensuring system performance, the scheduling algorithm can be simplified, energy saving and consumption reduction can be achieved, and services can be provided. System performance and reliability, specifically higher practicability.

The above are the preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, without departing from the principles of the present invention, several improvements and modifications can be made. These improvements and modifications It should also be regarded as the protection scope of the present invention.

Claims (11)

1. A scheduling method of a business system based on cloud computing is characterized by comprising the following steps:

when the business system switches a scene mode to a busy hour scene corresponding to a busy hour period, acquiring a mapping relation table which is stored in advance and is between a virtual machine and a physical server of the business system under the busy hour scene;

scheduling the virtual machine of the service system to a corresponding physical server according to the mapping relation in the mapping relation table in the busy hour scene;

when the business system switches a scene mode to an idle time scene corresponding to an idle time period, acquiring a mapping relation table between a virtual machine and a physical server of the business system under the idle time scene, wherein the mapping relation table is stored in advance;

scheduling the virtual machine of the service system to a corresponding physical server according to the mapping relation in the mapping relation table in the idle time scene;

the mapping relation table under the busy hour scene and the mapping relation table under the idle hour scene respectively record mapping relations between virtual machines and physical servers, which are obtained by the service system in the busy hour scene and the idle hour scene according to scheduling of a preset virtual machine scheduling algorithm, at the last time, and the busy hour scene and the idle hour scene respectively correspond to periodic busy hour time periods and idle hour time periods of the service system.

2. The method of claim 1,

when the business system is switched to a scene mode, if the mapping relation table in the scene mode is not stored, the resource occupation condition of each physical server in the business system is periodically collected, and the migration scheduling of the virtual machine is executed according to a preset virtual machine scheduling algorithm;

and recording and storing the mapping relation table in the scene mode according to the mapping relation between the virtual machine and the physical server after the migration scheduling.

3. The method of claim 1, wherein the scheduling the virtual machines of the business system to the corresponding physical servers according to the mapping relationship in the mapping relationship table comprises:

waking up the physical server recorded in the mapping relation table, and scheduling the virtual machine recorded in the mapping relation table to the corresponding physical server according to the mapping relation in the mapping relation table; and the number of the first and second groups,

and powering off or sleeping the rest physical servers in the service system except the physical servers recorded in the mapping relation table.

4. The method of claim 1, wherein after the step of scheduling the virtual machines of the business system to the corresponding physical servers according to the mapping in the mapping table, the method further comprises:

periodically collecting the resource occupation condition of each physical server in the service system, and executing migration scheduling of the virtual machine according to the resource occupation condition and a preset virtual machine scheduling algorithm;

and after the migration scheduling of the virtual machine is executed, updating the mapping relation table according to the mapping relation between the current virtual machine of the service system and the physical server.

5. The method of claim 2 or 4,

the working time period of the service system comprises busy hour time periods and idle time periods which are preset and periodically presented;

the executing migration scheduling of the virtual machine according to the resource occupation situation and a predetermined virtual machine scheduling algorithm comprises:

when the scene mode is a busy hour scene corresponding to the busy hour period, transferring a part of virtual machines on the physical server with the resource occupancy rate exceeding a preset threshold to the physical server with the lowest resource occupancy rate in the service system according to a load balancing algorithm;

and when the scene mode is an idle-time scene corresponding to the idle-time period, migrating the virtual machines of the service system to the physical servers with the least number according to a knapsack algorithm.

6. A scheduling apparatus of a cloud computing-based business system, comprising:

the system comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring a mapping relation table between a virtual machine and a physical server of the service system under a pre-stored busy hour scene when the service system switches a scene mode to the busy hour scene corresponding to a busy hour period, and acquiring the mapping relation table between the virtual machine and the physical server of the service system under a pre-stored idle hour scene when the service system switches the scene mode to the idle hour scene corresponding to the idle hour period;

the first scheduling unit is used for scheduling the virtual machine of the service system to a corresponding physical server according to the mapping relation in the mapping relation table under the busy hour scene, and scheduling the virtual machine of the service system to a corresponding physical server according to the mapping relation in the mapping relation table under the idle hour scene;

the mapping relation table under the busy hour scene and the mapping relation table under the idle hour scene respectively record mapping relations between virtual machines and physical servers, which are obtained by the service system in the busy hour scene and the idle hour scene according to scheduling of a preset virtual machine scheduling algorithm, at the last time, and the busy hour scene and the idle hour scene respectively correspond to periodic busy hour time periods and idle hour time periods of the service system.

7. The scheduling apparatus of claim 6, further comprising:

the second scheduling unit is used for periodically collecting the resource occupation condition of each physical server in the service system and executing migration scheduling of the virtual machine according to a preset virtual machine scheduling algorithm if the mapping relation table in the scene mode is not stored when the service system is switched to the scene mode;

and the first updating unit is used for recording and storing the mapping relation table in the scene mode according to the mapping relation between the virtual machine after the migration scheduling and the physical server.

8. The scheduling device as claimed in claim 7, wherein the working time period of the service system includes busy time periods and idle time periods which are preset and periodically presented; the second scheduling unit is further configured to migrate, according to a load balancing algorithm, a part of virtual machines on the physical server whose resource occupancy rate exceeds a predetermined threshold to the physical server with the lowest resource occupancy rate in the service system when the scene mode is a busy hour scene corresponding to the busy hour period; and when the scene mode is an idle-time scene corresponding to the idle-time period, migrating the virtual machines of the service system to the physical servers with the least number according to a knapsack algorithm.

9. The scheduling apparatus of claim 6, wherein the first scheduling unit comprises:

the migration unit is used for awakening the physical server recorded in the mapping relation table and scheduling the virtual machine recorded in the mapping relation table to the corresponding physical server according to the mapping relation in the mapping relation table; and the number of the first and second groups,

and the energy-saving unit is used for carrying out power-off or dormancy processing on the rest physical servers in the service system except the physical servers recorded in the mapping relation table.

10. The scheduling apparatus of claim 6, further comprising:

the third scheduling unit is used for periodically acquiring the resource occupation condition of each physical server in the service system and executing the migration scheduling of the virtual machine according to the resource occupation condition and a preset virtual machine scheduling algorithm; and after the migration scheduling of the virtual machine is executed, updating the mapping relation table according to the mapping relation between the current virtual machine of the service system and the physical server.

11. The scheduling apparatus as claimed in claim 10, wherein the operation time period of the service system comprises busy time period and idle time period which are preset and periodically presented;

the third scheduling unit is further configured to migrate, according to a load balancing algorithm, a part of virtual machines on the physical server whose resource occupancy rate exceeds a predetermined threshold to the physical server with the lowest resource occupancy rate in the service system when the scene mode is a busy hour scene corresponding to the busy hour period; and when the scene mode is an idle-time scene corresponding to the idle-time period, migrating the virtual machines of the service system to the physical servers with the least number according to a knapsack algorithm.

Images