JP4739272B2 - Load distribution apparatus, virtual server management system, load distribution method, and load distribution program - Google Patents

Description

  The present invention relates to a load distribution device, a virtual server management system, a load distribution method, and a load distribution program, and in particular, applied to a method of performing load distribution on a physical server based on the operating status of a virtual server operating on the physical server. Is preferred.

In order to execute a task that required multiple physical servers on one physical server, multiple virtual servers are operated on one physical server, and applications can be executed on each virtual server. Server virtualization technologies that enable use are becoming widespread. Also, with this server virtualization technology, since multiple virtual servers can be operated on one physical server, one or more physical servers can be flexibly adapted to changes in resource requirements. it is possible to shake Ri efficiently split and have the resources on the virtual server with.

  Also, for example, in Patent Document 1, in a CoD system that increases or decreases the number of servers according to the amount of requests from clients, a relay control device is provided between the clients and the CoD system to temporarily store received processing requests. At the same time, by controlling so that requests exceeding the current processing capacity do not flow into the CoD system, the system is overloaded due to sudden increase in the request amount, and consequently the response time is deteriorated, the request is discarded, the system is unstable. A method for avoiding the conversion is disclosed.

Further, for example, in Patent Document 2, when the amount of traffic indicated by the request amount prediction value arrives at the network service, the average response time for the user terminal is equal to or less than a response threshold predetermined by the operation manager. Thus, by controlling the amount of servers allocated to the network service, there is a method for automatically distributing the servers in the data center to each network service in real time to the load balancer without human intervention. It is disclosed.
JP-A-2005-250548 WO2004 / 092971

However, with the conventional server virtualization technology, the resources possessed by the physical server can be allocated on the virtual server in consideration of resource requirements, but the processing capacity and power consumption of the physical server, the number of allocated physical servers and abnormalities Considering various conditions such as degenerate capacity at the time, there is a problem that resources possessed by the physical server cannot be efficiently allocated on the virtual server.
Accordingly, an object of the present invention is to provide a load distribution device, a virtual server management system, a load distribution method, and a load that can efficiently allocate resources possessed by a physical server on a virtual server while considering various allocation conditions. To provide a distributed program.

In order to solve the above-described problem, according to the load distribution apparatus of claim 1, a load status collection unit that collects a load status of a system operated on a plurality of virtual servers from a plurality of physical servers, and the plurality Load allocation means for allocating the load applied to the physical server according to a genetic algorithm in which the load capacity of the system operated by the virtual server is expressed by a gene sequence, and the load allocated by the load allocation means is the physical server and a load distribution unit for distributing to the load allocating means, while setting the reference unit obtained by dividing the capacity of the processor of the physical server into a plurality as a single gene, running on the plurality of virtual servers Each system is assigned an individual symbol, and the load capacity of each system is divided by the reference unit. By arranging the same number of symbols, each of the plurality of systems is represented by a plurality of genes composed of the same symbols, and the number of symbols of each system is arranged according to the priority order of the plurality of systems. From the representation of the whole of the plurality of systems in one gene sequence, a gene is selected from a gene having an initial solid selection width using a random number according to an initial solid generation order to generate a plurality of initial solids. When the genes having the earliest order in the solid selection width are selected by the number of the system, the selection target is changed to the next gene in the initial solid selection width, and the number of genes assigned to the initial solid is the plurality of genes. Until the total number of genes in the system matches, the process after the process of generating the initial solid using the random number is repeated, and the initial When the number of genes assigned to the body matches the total number of genes of the plurality of systems, the solid genes obtained by repeating the selection and crossover according to the genetic algorithm from the plurality of initial individuals at that time are assigned. wherein the sequential assignment Rukoto to the physical servers within the load capacity set in the physical server in question.
The load distribution device according to claim 2, wherein the load allocating unit allocates the load so that the load applied to the physical server falls within a predetermined range of processing capacity of the physical server. To do.

According to a third aspect of the load distribution apparatus, when the load on the currently operating physical server exceeds the upper limit, the load on the physical server exceeds the upper limit. So that the load applied to the physical server is reassigned.
According to a fourth aspect of the present invention, the load allocating unit operates on the physical server on which the system is scheduled to stop when the physical server on which the system is operated on the virtual server is scheduled to be stopped. The system load is reassigned to another physical server.

According to the virtual server management system according to claim 5 , a plurality of physical servers, a plurality of virtual servers respectively started on the physical server, and a load capacity of a system operated on the virtual servers are arranged in a gene array according the represented genetic algorithm in the calculated load allocated to each physical server, the load the calculated and a load balancer to distribute the physical servers, the load balancer of the physical server While setting the reference unit obtained by dividing the processing capacity of the processor into a plurality of genes as one gene, each of the plurality of systems operated by the plurality of virtual servers is assigned with individual symbols, and the load on each system By arranging the number of the symbols based on the value obtained by dividing the capacity by the reference unit, that of the plurality of systems is arranged. This is expressed by a gene consisting of a plurality of the same symbols, and by arranging the number of symbols of each system according to the priority order of the plurality of systems, the whole of the plurality of systems is expressed by one gene sequence, Select a gene using a random number from the genes with the initial solid selection width according to the initial solid generation order to generate multiple initial solids, and select the genes with the earliest order in the initial solid selection width as many as the number of the system. Then, the order of selection in the initial solid selection width is changed to the next gene, and the initial number is selected using the random number until the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems. The processes after the process of generating the solid are repeated, and the number of genes assigned to the initial solid is equal to the total number of genes of the plurality of systems. In this case, a solid gene obtained by repeating the selection and crossover from a plurality of initial individuals at that time according to the genetic algorithm is transferred to the physical server within the range of the load capacity set for the physical server to be allocated. sequentially and wherein the allocation Rukoto.

Further, according to the load balancing method of claim 6, the step of collecting the load status of the system operated on the plurality of virtual servers from the plurality of physical servers, and the load capacity of the system operated on the virtual server include according to the genetic algorithm expressed by the gene sequence, a step of calculating the load to be allocated for each of the physical server, so as to perform load assigned to each of the physical servers and the step of dispersing the physical server, to a computer In the step of calculating the load assigned to each physical server, a reference unit obtained by dividing the processing capacity of the processor of the physical server into a plurality of sets is set as one gene, while the plurality of virtual servers In addition to assigning individual symbols to each of the multiple systems operating in By arranging the number of the symbols based on the value obtained by dividing the load capacity of the system by the reference unit, each of the plurality of systems is expressed by a gene composed of a plurality of the same symbols, and the priority of the plurality of systems is expressed. A gene using a random number among genes of an initial solid selection width according to an initial solid generation order from a plurality of systems expressed as a single gene sequence by arranging the number of symbols of each system according to rank To generate a plurality of initial solids, and when the genes having the earliest order in the initial solid selection width are selected by the number of the system, the selection target is changed to the next gene in the initial solid selection width. The initial solid is generated using the random number until the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems. When the number of genes assigned to the initial individual matches the total number of genes of the plurality of systems, the selection and selection are further performed from the plurality of initial individuals at that time according to a genetic algorithm. It is characterized in that the solid genes obtained by repetition are sequentially allocated to the physical servers within the range of the load capacity set for the physical servers to be allocated .

According to the load distribution program of claim 7, the step of collecting the load status of the system operated on the plurality of virtual servers from the plurality of physical servers, and the load capacity of the system operated on the virtual server include according to the genetic algorithm expressed by the gene sequence, which is executed a step of calculating the load to be allocated for each of the physical server, the steps of dispersing a load assigned to each of the physical servers on the physical server, to a computer In the step of calculating the load allocated to each physical server, a reference unit obtained by dividing the processing capacity of the processor of the physical server into a plurality of genes is set as one gene, while the plurality of virtual servers Individual symbols are assigned to each of a plurality of operating systems, and By arranging the number of symbols based on the value obtained by dividing the system load capacity by the reference unit, each of the plurality of systems is expressed by a gene consisting of a plurality of the same symbols, and the priority of the plurality of systems is expressed. A gene using a random number among genes of an initial solid selection width according to an initial solid generation order from a plurality of systems expressed as a single gene sequence by arranging the number of symbols of each system according to rank To generate a plurality of initial solids, and when the genes having the earliest order in the initial solid selection width are selected by the number of the system, the selection target is changed to the next gene in the initial solid selection width. The initial solid is generated using the random number until the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems. When the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems, the culling and crossover are further performed from the initial individuals at that time according to a genetic algorithm. It is characterized in that the solid genes obtained by repetition are sequentially allocated to the physical servers within the range of the load capacity set for the physical servers to be allocated .

  As described above, according to the present invention, according to a genetic algorithm, the load allocated to a system operated on a virtual server can be calculated for each physical server so as to satisfy a predetermined allocation condition. It is possible to efficiently allocate the resources possessed by the physical server on the virtual server while considering the conditions.

Hereinafter, a load distribution apparatus according to an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a schematic configuration of a virtual server management system to which the load distribution apparatus according to the first embodiment of the present invention is applied.
In FIG. 1, the virtual server management system is provided with n (n is an integer of 2 or more) physical servers 1a, 1b,. The physical server 1a is configured such that m (m is an integer of 2 or more) virtual servers 2a, 2b,..., 2m can be started, and the physical server 1b is configured with m virtual servers 3a and 3b. ,..., 3m can be started, the physical server 1c is configured to start m virtual servers 4a, 4b,..., 4m, and the physical server 1n is m virtual servers. 5a, 5b,..., 5m can be activated.

Note that the physical servers, referred to herein is configured to operate as a respective single server hardware, and virtual servers, configured on a physical server to operate as a respective one server by software It means what was done. An operating system such as Windows (registered trademark) can be individually started up on the virtual server. For this reason, by mounting a plurality of virtual servers on a physical server, a plurality of operating systems can be individually started up on one physical server.
And when it comprises so that a several virtual server can be started on one physical server, it can implement | achieve, for example by installing server virtualization software on a physical server.

  Further, the virtual server management system includes loads on the physical servers 1a, 1b,..., 1n that are applied to the virtual servers 2a, 2b,..., 2m, 3a, 3b,. .. 4m, 5a, 5b,..., A load distribution device 6 that distributes to 5m is provided. The load distribution device 6 includes virtual servers 2a, 2b, ..., 2m, 3a, 3b, ..., 3m, 4a, 4b, ..., 4m, 5a, 5b, ..., 5m. .., 1n that collects the load status of the system operated on the plurality of physical servers 1a, 1b,..., 1n, virtual servers 2a, 2b,..., 2m, 3a, 3b,. 3m, 4a, 4b,..., 4m, 5a, 5b,..., 5m. Load allocation means 8 for calculating the load allocated for each 1n, and load distribution means 9 for distributing the load calculated by the load allocation means 8 to the physical servers 1a, 1b,.

The virtual server referred to here may be an application that runs on a web server or a business-oriented system.
And the load condition collection means 7 is virtual server 2a, 2b, ..., 2m, 3a, 3b, ..., 3m, 4a, 4b, ..., 4m, 5a, 5b, ..., 5m. The load status of the system operated above is collected from a plurality of physical servers 1a, 1b,..., 1n, and the load status is notified to the load allocation means 8. And the load allocation means 8 is virtual server 2a, 2b, ..., 2m, 3a, 3b, ..., 3m, 4a, 4b, ..., 4m, 5a, 5b, ..., 5m. In accordance with a genetic algorithm in which the load capacity of the system to be operated is expressed by a gene sequence, the load applied to the physical servers 1a, 1b..., 1n is allocated, and the allocation result is notified to the load distribution means 9. The load distribution unit 9 can distribute the load allocated by the load allocation unit 8 to the physical servers 1a, 1b,.

That is, when allocating the load applied to the physical servers 1a, 1b,..., 1n, the load allocating means 8 uses the reference unit obtained by dividing the processing capacity of the processor as one gene, and the virtual servers 2a, 2b,. 2m, 3a, 3b, ..., 3m, 4a, 4b, ..., 4m, 5a, 5b, ... Generates a solid in which the load capacity of the system operating at 5m is expressed by gene sequences can do.
Then, the solid wrinkles and crossovers are repeated according to the genetic algorithm, and the solid gene obtained at that time is assigned to the physical server within the range of the load capacity set in the physical servers 1a, 1b,. 1a, 1b,..., 1n can be assigned sequentially.

Then, when a solid gene is assigned to the physical servers 1a, 1b,..., 1n, based on the solid genes assigned to the physical servers 1a, 1b. ... 2m, 3a, 3b, ... 3m, 4a, 4b, ... 4m, 5a, 5b, ... Loads assigned to the system operating at 5m are the physical servers 1a, 1b ... can be calculated every 1n.
Then, solid dredging and crossover can be repeated according to the genetic algorithm until the load allocated to the system for each physical server 1a, 1b,..., 1n satisfies a predetermined allocation condition.

When the load applied to the physical servers 1a, 1b,..., 1n is allocated by the load allocation unit 8, the load distribution unit 9 converts the load calculated by the load allocation unit 8 into the physical servers 1a, 1b,. -It can be distributed to 1n. The physical servers 1a, 1b,..., 1n are virtual servers 2a, 2b, ..., 2m, 3a, 3b, ..., 3m, 4a, according to the load calculated by the load assigning means 8. The system can be operated on 4b, ..., 4m, 5a, 5b, ..., 5m.
As a condition for assigning the load on the physical servers 1a, 1b,..., 1n, for example, the load on the physical servers 1a, 1b. The load can be allocated so as to fall within a predetermined range.

Alternatively, when the physical servers 1a, 1b,..., 1n are operating, for example, when the load on the physical server 1a exceeds the upper limit value, the load on the physical server 1a does not exceed the upper limit value. In addition, the load on the physical servers 1a, 1b,.
Alternatively, for example, when there is a plan to stop the physical server 1a on which the system is operated on the virtual servers 2a, 2b,..., 2m, the load on the system operated on the physical server 1a is set to another physical server 1b. ... can be reassigned to 1n.

  Thus, according to the genetic algorithm, the virtual servers 2a, 2b, ..., 2m, 3a, 3b, ..., 3m, 4a, 4b, ..., 4m, 5a, 5b, ..., 5m The load allocated to the system to be operated can be calculated for each of the physical servers 1a, 1b,..., 1n so as to satisfy a predetermined allocation condition, and the physical servers 1a, 1b,. .., 2m, 3a, 3b,..., 3m, 4a, 4b,..., 4m, 5a, 5b,. It becomes possible to allocate efficiently over 5m.

FIG. 2 is a diagram showing an example of a method for allocating a system operating on a virtual server according to an embodiment of the present invention onto a physical server.
In FIG. 2, for example, the A system of company A is operated on the physical server 1a of FIG. 1, the b system of company B is operated on the physical server 1b of FIG. 1, and the c system of company C is the physical system of FIG. It is assumed that it is operating on the server 1c.

  1 collects the load capacities on the physical servers 1a, 1b, and 1c of company A's system a, company B's system b, and company C's system c in time units or days. be able to. The load capacity of company A's system a on the physical server 1a is high in the first half of the month, and the load capacity of company B's system b on the physical server 1b is high in the middle of the month. Assuming that the load capacity of the c system is high in the second half of the month, the load allocation device 6 allocates the load capacity of the A system of company A, the b system of company B and the c system of company C to one physical server 1a. The load capacity of the A system of company A, the b system of company B, and the c system of company C can be borne by one physical server 1a.

  When the load capacities of company A's a system, company B's b system, and company C's c system are allocated to physical server 1a, physical server 1a starts virtual servers 2a, 2b, and 2c, and company A's a The system can be operated on the virtual server 2a, the B system of company B can be operated on the virtual server 2b, and the c system of company C can be operated on the virtual server 2c.

FIG. 3 is a block diagram showing a schematic configuration of a virtual server management system to which the load distribution apparatus according to the second embodiment of the present invention is applied.
3, the virtual server management system is provided with n physical servers 11a, 11b,. The physical server 11a is configured to start m virtual servers 12a, 12b,..., 12m, and the physical server 11b starts m virtual servers 13a, 13b,. The physical server 11c is configured to start m virtual servers 14a, 14b,..., 14m, and the physical server 11n is configured to m virtual servers 15a, 15b,. 15m can be activated.

  The physical server 11a includes a virtual machine management manager 16, an operation status collection unit 17, an operation result transmission unit 18, an operation result data 19, an operation result collection unit 20, a virtual machine control unit 21, and a virtual machine power control instruction unit 22. Is provided. Although not shown in FIG. 3, the virtual machine management manager 16, the operation status collection unit 17, the operation result transmission unit 18, and the operation are performed on the other physical servers 11 b... 11 n as well as the physical server 11 a. Achievement data 19, operation achievement collection means 20, and virtual machine control means 21 can be provided.

  Here, the virtual machine management manager 16 manages the load capacity of the virtual servers 12a, 12b,..., 12m, and instructs the physical server 11a to start and stop the virtual servers 12a, 12b,. You can do it. The operation status collection means 17 can collect the operation status of the physical server 11a and the virtual servers 12a, 12b,..., 12m, and send a request for transmission of the operation status to the load balancer 31. The operation result transmission means 18 can periodically make a request to transmit the operation results of the physical server 11a and the virtual servers 12a, 12b,.

  The operation result collection means 20 can collect the operation results of the physical server 11a and the virtual servers 12a, 12b,..., 12m and generate the operation result data 19. The virtual machine control means 21 controls the start and stop of the virtual servers 12a, 12b,..., 12m based on the load assigned to the physical server 11a and the virtual servers 12a, 12b,. The manager 16 can be instructed. The transmission / reception control unit 23 can control transmission / reception of data between the physical server 11 a and the load distribution device 31.

Further, the operation result data 19 indicates the operation results in the system unit of the physical server 11a and the virtual servers 12a, 12b,..., 12m. be able to.
-Virtual server name key-CPU load for each system when the physical server 11a and the virtual servers 12a, 12b, ..., 12m are operated-The physical server 11a and the virtual servers 12a, 12b, ..., 12m Memory usage per system when it is running

  Note that the virtual machine management manager 16, the operation status collection unit 17, the operation result transmission unit 18, the operation result collection unit 20, and the virtual machine control unit 21 execute a program in which an instruction for performing processing performed by these units is described. This can be realized by causing a computer to execute it. The operation result data 19 can be stored in a memory in the physical server 11a.

  Then, if this program is stored in a storage medium such as a CD-ROM, the virtual machine management manager 16 can collect the operation status by installing the storage medium in the computer of the physical server 11a and installing the program in the computer. The processing performed by the means 17, the operation result transmission means 18, the operation result collection means 20, and the virtual machine control means 21 can be realized.

  In addition, the virtual server management system is provided with a load balancer 31. The load balancer 31 includes virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ... .., 14m, 15b,..., 15m are allocated to the physical servers 11a, 11b,. , 11n can be instructed to power off the physical servers 11a, 11b,..., 11n based on the load assigned to 11n.

  The load distribution device 31 includes a transmission / reception control unit 32, a load distribution device 31, an operation status reception unit 33, an operation result reception unit 34, a virtual machine state table 35, operation result data 36, a virtual machine abnormality monitoring unit 37, a physical A machine management table 38, an allocation system management table 39, a virtual machine allocation scheduler 40, hourly operation result data 41, virtual machine switching schedule data 42, virtual machine control instruction means 43, and load distribution means 45 are provided.

  Here, the transmission / reception control unit 32 can control transmission / reception of data between the physical server 11 a and the load distribution apparatus 31. The operating state receiving means 33 includes physical servers 11a, 11b,..., 11n, and virtual servers 12a, 12b,..., 12m, 13a, 13b, ..., 13m, 14a, 14b,. The operating states of 15a, 15b,..., 15m can be received and registered in the virtual machine state table 35.

The operation result receiving means 34 includes the physical servers 11a, 11b,..., 11n and the virtual servers 12a, 12b,..., 12m, 13a, 13b, ..., 13m, 14a, 14b,. 15a, 15b,..., 15m operation results can be received and registered as operation result data 36.
The virtual machine abnormality monitoring means 37 refers to the virtual machine state table 35 periodically, while referring to the virtual servers 12a, 12b,..., 12m, 13a, 13b, ..., 13m, 14a, 14b,. , 14m, 15a, 15b,..., 15m, abnormalities in operating conditions, protrusion of load capacity, etc. are detected, and the virtual servers 12a, 12b,..., 12m, 13a, 13b,. , 14a, 14b,..., 14m, 15a, 15b,..., 15m can be reset.

  The virtual machine allocation scheduler 40 operates the virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m. .., 12m, 13a, 13b,..., 13m, 14a, 14b,..., 14m, 15a, 15b,. A schedule can be set.

  Here, the virtual machine allocation scheduler 40 is the virtual servers 12a, 12b,..., 12m, 13a, 13b,..., 13m, 14a, 14b, ..., 14m, 15a, 15b,. When setting a load allocation schedule for 15m, virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ... , 12m, 13a, 13b,..., 13m, 14a, 14b,... According to a genetic algorithm in which the load capacity of the system operating at 15 m is expressed by a gene sequence. ., 14m, 15a, 15b,..., 15m can be assigned loads.

  The virtual machine control instruction means 43 periodically monitors the virtual machine switching schedule data 42, and physical servers 11a, 11b,..., 11n and virtual servers 12a, 12b,..., 12m, 13a, 13b,. ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., based on the load allocated to 15m, those physical servers 11a, 11b ..., 11n and virtual servers 12a, 12b , ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m so that the physical servers 11a, 11b,.・ You can instruct 11n.

Based on the virtual machine switching schedule data 42, the load distribution means 45 is based on the virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, The load can be distributed to 15b, ..., 15m.
The virtual machine state table 35 includes virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m. The management information of each of the above systems is shown, and examples of management items in the virtual machine state table 35 include the following.
-Virtual server name key-Operating status of each system-Status code of each system-Status message contents of each system-Details of status messages of each system

The operation result data 36 includes physical servers 11a, 11b ..., 11n and virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a. , 15b,..., 15m indicates the operation results in units of systems, and examples of management items in the operation result data 36 include the following.
Virtual server name key Physical servers 11a, 11b ..., 11n and virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a , 15b,..., 15m when the CPU load per system unit / memory usage rate per system unit

The physical machine management table 38 shows the management information of the physical servers 11a, 11b,..., 11n on which the system is actually operated, and examples of management items in the physical machine management table 38 include the following. it can.
-Physical server name key-Physical server operation status-Physical server allocation priority order-Physical server status code-Physical server status message content-Details of physical server status message-CPU capacity of physical server (machine specification-current Capacity state)
・ Memory capacity of physical server (machine spec / current capacity status)

The allocation system management table 39 is on the virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m. The management information of the system to be operated is shown. Examples of management items in the allocation system management table 39 include the following examples.
-Application system unit key-Company code and company name-System name-Schedule pattern (previous week, previous month, previous year, etc.)
・ Maximum CPU load of each system ・ Maximum memory capacity of each system ・ CPU threshold value for each system ・ Memory threshold value for each system ・ Maximum transaction time of each system (stop waiting time)
-Whether each system can be forcibly disconnected-Whether there is a notification when each system is stopped-Message contents when each system is stopped-Physical server that cannot be started

.., 11n and virtual servers 12a, 12b,..., 12m, 13a, 13b,..., 13m, 14a, 14b,. , 15a, 15b,..., 15m indicates the operation results for each hour, and examples of management items in the hourly operation results data 41 include the following examples.
-Application system unit key-Company code and company name-System name-Date and time of system operation-Physical servers 11a, 11b ..., 11n and virtual servers 12a, 12b, ..., 12m, 13a, 13b,..., 13m, 14a, 14b,..., 14m, 15a, 15b,..., 15m when the maximum CPU load in the system unit, the maximum memory usage rate in the system unit, Average CPU load per system / Average memory usage per system

The virtual machine switching schedule data 42 includes which virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m. The above shows the operation plan data about when to execute which system while allocating how much load capacity, and examples of management items in the virtual machine switching schedule data 42 include the following.
-Virtual server key-Virtual server start / stop date-CPU capacity at virtual server startup-Memory capacity at virtual server startup

  The transmission / reception control means 23, the transmission / reception control means 32, the load balancer 31, the operation status reception means 33, the operation result reception means 34, the virtual machine abnormality monitoring means 37, the virtual machine allocation scheduler 40, the virtual machine control instruction means 43, the load The distribution unit 45 can be realized by causing a computer to execute a program in which an instruction for performing processing performed by these units is described. Further, the virtual machine state table 35, operation result data 36, physical machine management table 38, allocation system management table 39, hourly operation result data 41, and virtual machine switching schedule data 42 are stored in a memory in the load balancer 31. be able to.

  If this program is stored in a storage medium such as a CD-ROM, the transmission / reception control means 23, the transmission / reception control means are installed by installing the storage medium in the computer of the load distribution apparatus 31 and installing the program in the computer. 32, the load distribution device 31, the operation state reception unit 33, the operation result reception unit 34, the virtual machine abnormality monitoring unit 37, the virtual machine allocation scheduler 40, the virtual machine control instruction unit 43, and the processing performed by the load distribution unit 45 are realized. be able to.

  In the physical server 11a, the operation status collection unit 17 collects the operation status of the physical server 11a and the virtual servers 12a, 12b,..., 12m, and distributes the operation status via the transmission / reception control unit 23. Transmit to device 31. Further, the operation result collecting means 20 collects the operation results of the physical server 11a and the virtual servers 12a, 12b,..., 12m, and holds the operation results as operation result data 19 in the physical server 11a. When the operation result data 19 is stored in the physical server 11 a, the operation result transmission unit 18 transmits the operation result data 19 to the load distribution device 31 via the transmission / reception control unit 23.

The other physical servers 11b... 11n can operate in the same manner as the physical server 11a.
Then, when the operating statuses of the physical servers 11a, 11b,..., 11n are transmitted to the load balancer 31, the operating status receiving means 33 operates the physical servers 11a, 11b,. The situation is received via the transmission / reception control means 33 and registered in the virtual machine state table 35.
When the operation results of the physical servers 11a, 11b,..., 11n are respectively transmitted to the load balancer 31, the operation result receiving unit 34 operates the physical servers 11a, 11b,. The result is received via the transmission / reception control means 33 and is held as operation result data 36.

  Then, the virtual machine abnormality monitoring means 37 periodically refers to the virtual machine state table 35 and the operation result data 36, and the virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b,..., 14m, 15a, 15b,. And the virtual machine abnormality monitoring means 37 is the virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ...,. The operation schedule is reset according to the abnormality of the 15 m operation status or the protrusion of the load capacity, and the virtual machine switching schedule data 42 is held.

  Further, the virtual machine allocation scheduler 40 refers to the physical machine management table 38 and the allocation system management table 39, and based on the operation result data 36, the virtual servers 12a, 12b,..., 12m, 13a, 13b,. .., 13m, 14a, 14b,..., 14m, 15a, 15b,.

  Here, the virtual machine allocation scheduler 40 is the virtual servers 12a, 12b,..., 12m, 13a, 13b,..., 13m, 14a, 14b, ..., 14m, 15a, 15b,. When setting a load allocation schedule for 15 m, the virtual servers 2 a, 2 b,..., 2 m, 3 a, 3 b,. 4b,..., 4m, 5a, 5b,..., 5m can generate a solid in which the load capacity of the system is expressed by gene sequences.

Then, the virtual machine allocation scheduler 40 repeats solid selection and crossover according to a genetic algorithm, and refers to the physical servers 1a, 1b,. Can be allocated sequentially.
When the virtual machine allocation scheduler 40 allocates a solid gene to the physical servers 1a, 1b,..., 1n, based on the solid genes allocated to those physical servers 1a, 1b. .. 2m, 3a, 3b,..., 3m, 4a, 4b,..., 4m, 5a, 5b,. Can be calculated for each physical server 1a, 1b,.

Then, the virtual machine allocation scheduler 40 repeats solid selection and crossover according to a genetic algorithm until the load allocated to the system for each physical server 1a, 1b, ..., 1n satisfies a predetermined allocation condition. Can do.
The virtual machine control instruction means 43 periodically monitors the virtual machine switching schedule data 42, and physical servers 11a, 11b,..., 11n and virtual servers 12a, 12b,..., 12m, 13a, 13b, .., 13m, 14a, 14b,..., 14m, 15a, 15b,..., 15m, and the physical servers 11a, 11b,. , 12b,..., 12m, 13a, 13b,..., 13m, 14a, 14b,..., 14m, 15a, 15b,. Through the physical servers 11a, 11b,.

  For example, when an instruction from the virtual machine control instruction unit 43 is given to the physical server 11a, the virtual machine control unit 21 receives an instruction from the virtual machine control instruction unit 43 via the transmission / reception control unit 23, and the physical server The virtual machine management manager 16 is instructed to start and stop the virtual servers 12a, 12b,..., 12m based on the loads assigned to the virtual servers 11a and 12a, 12b,.

Further, the load distribution means 45 is based on the virtual machine switching schedule data 42, and the virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, The load is distributed to 15a, 15b, ..., 15m.
For example, when the load is distributed to the physical server 11a by the load distribution means 45, the virtual machine management manager 16 manages the load capacity of the virtual servers 12a, 12b,. , 12b,..., 12m are instructed to the physical server 11a.

FIG. 4 is a flowchart illustrating a method for allocating a system operated on a virtual server using a genetic algorithm according to an embodiment of the present invention onto a physical server.
In FIG. 4, when assigning a load by a genetic algorithm, after generating an initial solid (step S1), the system is arranged on the physical servers 11a, 11b,..., 11n (step S2).
Then, while evaluating the solid generated by the genetic algorithm and storing the excellent solution (step S3), until the predetermined allocation condition is satisfied (step S7), the solid wrinkles, crossover and mutation are performed. It can be repeated (steps S4 to S6).

  Here, when generating a solid in the genetic algorithm, a virtual unit 12a, 12b,..., 12m, 13a, 13b,. 14a, 14b,..., 14m, 15a, 15b,..., 15m can be a solid gene sequence allocated according to the load capacity of the system.

Note that the number of genes per individual can be managed as a system constant. In addition, a plurality of the same genes may exist in one solid, or a plurality of solids may exist in the system.
For example, assuming that the CPU capacity of an Intel Pentium (registered trademark) processor operating at a frequency of 3 GHz is 100, the CPU capacity obtained by dividing the CPU capacity by 10 can be used as one gene. Then, a unique gene can be generated for each system to which a load is assigned, and a number of genes corresponding to the load capacity of the system can be allocated to the system.

Here, in the production of the initial solid, specifically, the following procedure can be taken.
(1) The load capacity of a system to be assigned a load is divided for each CPU capacity of one gene, and genes unique to the system are assigned by the number of divisions.
(2) The genes are arranged according to the priority order of the system and the priority order of the physical servers 11a, 11b,.
(3) Except for the solid generated by the method of (2), the initial solid is generated using the initial solid selection width and the random number. In particular,
a) A gene is selected using a random number from genes having an initial solid selection width according to the initial solid production order.
b) When the genes with the earliest order in the initial solid selection width are selected by the number of the systems, the selection target gene is changed to the next gene.
c) The processes of a) and b) are repeated until the number of genes assigned to the solid matches the total number of genes of all systems to which the load is assigned.

  On the other hand, it is assumed that five physical servers 11a, 11b,..., 11e are set as initial conditions for the physical servers 11a, 11b,. The spec capacity of the physical server 11a is 280, the spec capacity of the physical server 11b is 190, the spec capacity of the physical server 11c is 440, the spec capacity of the physical server 11d is 70, and the spec capacity of the physical server 11e is 170. Assume that the initial value of the load capacity allocated to each of the physical servers 11a, 11b,..., 11e is 70% of the spec capacity.

In this case, the load capacity assignable to the physical server 11a is 196≈190, the load capacity assignable to the physical server 11b is 133≈130, the load capacity assignable to the physical server 11c is 308≈300, and the load capacity assignable to the physical server 11d. The possible load capacity is 49≈40, and the load capacity that can be allocated to the physical server 11e is 119≈110.
If the priority order of the physical servers 11a, 11b,..., 11e at the time of load capacity allocation is 2, 3, 1, 4, 5, respectively, the physical server 11c → the physical server 11a → the physical server 11b → physical Loads can be assigned in the order of server 11d → physical server 11e.

For example, it is assumed that there are four systems, ie, Company A system a, Company A system b, Company C system c, and Company D system, and the load capacity of the company A system is 80, Assume that the load capacity of company A's b system is 140, the load capacity of company C's c system is 200, and the load capacity of company D's d system is 80.
In this case, a gene “a” is assigned to the A company a system, and the load capacity of the A company a system can be expressed by eight genes “a”. Further, the gene “b” can be assigned to the company B system, and the load capacity of the company B system can be expressed by 14 genes “b”. Further, a gene “c” can be assigned to the C company C system, and the load capacity of the C company c system can be expressed by 20 genes “c”. Further, a gene “d” is assigned to the D company d system, and the load capacity of the D company d system can be expressed by eight genes “d”.

Then, the genes of the A company a system, the A company b system, the C company c system, and the D company d system according to the priority order of the systems are equal to the number of each system (A company a system = 8, A company b system = 14 By arranging them one by one, C company c system = 20, D company d system = 8), it is possible to generate 50 gene sequences “aaaaaaaaabbbbbbbbbbbbbbbbccccccccccccccccccccccdddddd”.
If the solid length is 10, by selecting 10 genes from 50 gene sequences of “aaaaaaaaabbbbbbbbbbbbbbbbbbccccccccccccccccccccccdddddd” according to the methods of (3) a), b) and c), 5 solids K1, K2, K3, K4, K5 can be produced as initial solids.

  If the initial solid generation order is “abcd” and the initial solid selection width is 3, “abc” is designated as the gene to be selected, and 10 genes are selected from the gene “abc” using random numbers. By selecting, for example, ten gene sequences “caabbcbcaa” can be generated as the initial solid K1. Further, by selecting 10 genes from the gene “abc” using random numbers, for example, 10 gene sequences “acabacbbba” can be generated as the initial solid K2. Here, in the solids K1 and K2, since eight genes “a” given to the A company a system are selected, the gene to be selected in the initial solid selection width is changed to the next gene, and the selection target “Bcd” is designated as the gene for.

  Then, by selecting 10 genes from the gene “bcd” using random numbers, for example, 10 gene sequences “dcccdbbcbd” can be generated as the initial solid K3. Further, by selecting 10 genes from the gene “bcd” using random numbers, for example, 10 gene sequences “dccdbccccd” can be generated as the initial solid K4. Further, by selecting 10 genes from the gene “bcd” using random numbers, for example, 10 gene sequences “cbbbbbbbbd” can be generated as the initial solid K5.

The following gene sequences can be generated as solid K1, K2, K3, K4, and K5 in a generation by repeating solid wrinkles, crossovers, and mutations according to a genetic algorithm.
Solid K1 = "cdbbabbccca"
Solid K2 = "ccbbccabcd"
Solid K3 = “acdbbbacc”
Solid K4 = “bbcabdbccca”
Solid K5 = "cccabcbccdd"
Then, when solids K1, K2, K3, K4, and K5 in a generation are generated, the physical servers 11a, 11b,..., 11e are ordered according to the priority order of the physical servers 11a, 11b,. The genes of the solid K1, K2, K3, K4, and K5 can be sequentially allocated to the physical servers 11a, 11b,..., 11e so that the load capacity that can be allocated is reached.

  For example, since the first place in the above-described assignment priority order is the physical server 11c, the physical servers 11c, K2, K3, K4, and K5 are sequentially assigned until the load capacity ≈300 that can be assigned to the physical server 11c is reached. The gene assigned to 11c is “cddbbabbccaccbbbccabcdaccdbbaccacc”. Then, when the genes of the solid K1, K2, K3, K4, and K5 are allocated to the physical server 11c, the physical server 11a with the second highest allocation priority is selected. When the remaining solid K4 and K5 genes are sequentially assigned until the load capacity ≈ 190 that can be assigned to the physical server 11a is reached, the gene assigned to the physical server 11a becomes “bbcabdbccaccccbcbcd”. Then, when the solid K4 and K5 genes are allocated to the physical server 11a, the physical server 11b having the third highest allocation priority is selected. When the gene of the remaining solid K5 is assigned to the physical server 11b, the gene assigned to the physical server 11b is “d”.

When the genes assigned to the physical servers 11c, 11a, and 11b are rearranged in units of systems, the physical server 11c → “aaaaaabbbbbbbbccccccccccccccdddd”, the physical server 11a → “aaabbbbbbbccccccccdd”, and the physical server 11b → “d”.
As a result, 5 genes “a”, 8 genes “b”, 12 genes “c”, and 5 genes “d” are allocated to the physical server 11c. 50 CPU capacity is allocated to the A company a system, 80 CPU capacity is allocated to the A company b system, 120 CPU capacity is allocated to the C company c system, and 50 CPU capacity is allocated to the D company d system. .

  In addition, three genes “a”, six genes “b”, eight genes “c”, and two genes “d” are allocated to the physical server 11a. 30 CPU capacities are allocated to the company a system, 60 CPU capacities are allocated to the company A b system, 80 CPU capacities are allocated to the C company c system, and 20 CPU capacities are allocated to the D company d system.

Further, one gene “d” is allocated to the physical server 11b, and 10 CPU capacities are allocated to the D company d system on the physical server 11b.
When the system is arranged on the physical servers 11a, 11b,..., 11n, it is determined whether or not the CPU capacity allocated to the physical servers 11a, 11b. To do. Here, for example, the load capacity that can be allocated to each of the physical servers 11a, 11b,..., 11n can be set to 80% as the upper limit, and the target load capacity that can be allocated can be set to 70%.

FIG. 5 is a flowchart illustrating a method for allocating a physical server load to a system operated on a virtual server according to an embodiment of the present invention.
5, in the implementation of the allocation to the physical servers 11a, 11b,..., 11n in step S2 in FIG. 4, the target load capacity calculation process (step S11) and the physical servers 11a, 11b,. A load capacity allocation process to 11n (step S12) is performed.

  In the target load capacity calculation process, the load capacity and average allocated capacity previously assigned to each physical server 11a, 11b,..., 11n are acquired (step S21). When any of the load capacities allocated to the physical servers 11a, 11b,..., 11n is 50% or less and the average allocated capacity of the physical servers 11a, 11b,. ), The number of physical servers 11a, 11b,. (Step S31).

  On the other hand, any of the load capacities allocated to the physical servers 11a, 11b..., 11n does not satisfy the condition that the average allocated capacity of the physical servers 11a, 11b. In the case (step S22), it is determined whether or not the load capacity of the allocated system has overflowed with respect to the number of physical servers 11a, 11b,..., 11n to be allocated (step S25).

  When the load capacity of the system to be allocated overflows with respect to the number of physical servers 11a, 11b,..., 11n to be allocated, one physical server 11a, 11b,. After increasing by the amount (step S23), the target load capacity that can be allocated is calculated the most (step S24), and the process proceeds to the next process (step S31).

  On the other hand, when the load capacity of the system to be allocated does not overflow with respect to the number of physical servers 11a, 11b,..., 11n to be allocated, the load allocated to the physical servers 11a, 11b. It is determined whether or not the capacity is smaller than the target value (step S27), and if the load capacity allocated to the physical servers 11a, 11b,. (Step S28), the process proceeds to the next process (step S31).

On the other hand, when the load capacity allocated to the physical servers 11a, 11b,..., 11n is larger than the target value (step S29), the target allocatable load capacity is increased by 1% (step S30), and then The process proceeds (step S31).
Then, when a load is allocated to the physical servers 11a, 11b,..., 11n, the solid is evaluated according to a predetermined allocation condition. Here, examples of the load allocation conditions for the physical servers 11a, 11b,..., 11n include the following items.
The number of system divisions is set to 2 or more in consideration of the response at the time of abnormality of the physical servers 11a, 11b,. This condition is an absolute condition and can be applied on the genetic algorithm.

.., 12m, 13a, 13b,..., 13m, 14a, 14b,..., 14m, 15a, 15b,. The number of divisions is 3 → 4 → 2 → 5 → 6 →. This condition is an arbitrary condition and can be applied on the genetic algorithm.
In consideration of power saving, increase the priority order of the physical servers 11a, 11b,. This condition is an arbitrary condition and can be applied in the pre-processing of the genetic algorithm.

Each system allocates equally to the physical servers 11a, 11b,..., 11n in order to secure a degenerate capacity when the physical servers 11a, 11b,. This condition is an arbitrary condition and can be applied on the genetic algorithm.
-Do not exceed the total amount of memory used by each physical server 11a, 11b, ..., 11n. This condition is an absolute condition and can be applied on the genetic algorithm.
The physical servers 11a, 11b,..., 11n scheduled for maintenance are excluded from allocation targets. This condition is an absolute condition and can be applied in the preprocessing of the genetic algorithm.

In order to equalize the margin, the CPU capacity of the physical servers 11a, 11b,. This condition is an arbitrary condition and can be applied on the genetic algorithm.
In consideration of power saving, the total power consumption of the physical servers 11a, 11b,. This condition is an arbitrary condition and can be applied on the genetic algorithm.

FIG. 6 is a flowchart illustrating a method for allocating a physical server load to a system operating on a virtual server when the physical server is overloaded according to an embodiment of the present invention.
In FIG. 6, the virtual machine abnormality monitoring means 37 in FIG. 3 waits for a fixed time (step S41), then selects the next company according to the priority order (step S42), and further selects the next system according to the priority order (step S42). Step S43). When the virtual machine abnormality monitoring unit 37 selects the next company and the next system, the virtual machine abnormality monitoring unit 37 calculates the load capacity for M minutes of the system (step S44). Then, if the N% CPU threshold value registered in the allocation system management table 39 continues for M minutes or more (step S45), the amount of increase in load allocated to the system is calculated (step S46). A certain virtual server 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m is acquired (step S47).

And the acquired virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m The load is allocated (step S48) and written in the virtual machine switching schedule data 42 (step S49).
Then, the virtual machine abnormality monitoring unit 37 waits for a certain time (step S50), and then, from the virtual machine switching schedule data 42, the virtual servers 12a, 12b,..., 12m, 13a, 13b,. , 14a, 14b,..., 14m, 15a, 15b,..., 15m are acquired (step S51). Then, while giving priority to an immediate operation request, if there is one whose start / stop time has passed (step S52), this is notified to the virtual machine control instruction means 43 (step S53).

FIG. 7 is a flowchart showing a method for allocating a physical server load to a system operating on a virtual server when a physical server or a virtual server is abnormal according to an embodiment of the present invention.
7, the virtual machine abnormality monitoring means 37 in FIG. 3 includes physical servers 11a, 11b,..., 11n or virtual servers 12a, 12b,..., 12m, 13a, 13b,. 14b,..., 14m, 15a, 15b,..., 15m are detected (step S151), it is determined whether the abnormality has occurred in the physical servers 11a, 11b. Step S152).

  When the abnormality has occurred in the physical servers 11a, 11b,..., 11n, the virtual machine abnormality monitoring means 37 has all the virtual machines operating in the physical servers 11a, 11b,. , 12m, 13a, 13b,..., 13m, 14a, 14b,..., 14m, 15a, 15b,..., 15m are collected (step S154). .

  On the other hand, the abnormality occurred in the virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m , The virtual servers 12a, 12b,..., 12m, 13a, 13b,..., 13m, 14a, 14b,..., 14m, 15a, 15b,. ... Collects a load capacity of 15 m (step S153).

  Then, the physical servers 11a, 11b,..., 11n or the virtual servers 12a, 12b,..., 12m, 13a, 13b, ..., 13m, 14a, 14b,. , 15b,..., 15m is calculated (step S155), and free virtual servers 12a, 12b,..., 12m, 13a, 13b,. .., 14m, 15a, 15b,..., 15m are acquired (step S156).

And the acquired virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m The load is allocated (step S157) and written to the virtual machine switching schedule data 42 (step S158).
Then, the virtual machine abnormality monitoring means 37 waits for a certain time (step S60), and then, from the virtual machine switching schedule data 42, the virtual servers 12a, 12b,..., 12m, 13a, 13b,. , 14a, 14b,..., 14m, 15a, 15b,..., 15m are acquired (step S61). Then, when priority is given to an operation that has an immediate request while there is an operation whose start / stop time has passed (step S62), the virtual machine control instruction unit 43 is notified of this (step S63).

FIG. 8 is a diagram showing a method of allocating a physical server load to a system operating on a virtual server when the physical server maintenance transition or a virtual server abnormality according to an embodiment of the present invention is performed. .
In FIG. 8, in the maintenance transition process of the physical servers 11a, 11b,. At time t1, acceptance of a new request for the physical servers 11a, 11b,..., 11n scheduled for maintenance is stopped, and at time t2, the physical servers 11a, 11b,. Notification is made, and the physical servers 11a, 11b,..., 11n scheduled for maintenance are stopped at time t3.

  For example, the time from the scheduled maintenance start time to the terminal notification start time can be 10 minutes. Further, if the A system of Company A is operating on the physical server 11a scheduled for maintenance, and the maximum transaction time of the A system of Company A is 25 minutes, the time from the scheduled maintenance start time to the cancellation of new request acceptance is 25. The time from the scheduled maintenance start time to the alternative server activation process can be 1 hour and 25 minutes.

Then, when the alternative server activation process is performed in accordance with the virtual machine switching schedule data 42 in FIG. 3 and the acceptance of a new request to the physical server 11a scheduled for maintenance is stopped, the virtual server 12a operating on the physical server 11a. , 12b,..., 12m decrease sequentially.
Then, an alarm is notified to the client terminal operating at the terminal notification start time by a pop-up message, and when the scheduled maintenance start time is reached, the maintenance scheduled physical server 11a is stopped. Here, when processing is being performed on the physical server 11a, a message can be notified to the administrator, and the processing can be suspended or forcibly stopped.

FIG. 9 is a flowchart showing an operation at the time of stopping acceptance of a new request at the time of maintenance shift of the physical server according to an embodiment of the present invention.
9, the virtual machine abnormality monitoring means 37 in FIG. 3 reads the operation schedule of the physical servers 11a, 11b..., 11n scheduled for maintenance from the allocation system management table 39 (step S71), and (maintenance start time- It is determined whether or not (maximum transaction time) <current time (step S72).

  If (maintenance start scheduled time−maximum transaction time) <current time, the item “machine state” in the physical machine management table 38 is changed to “maintenance stop requested” (step S73). Then, when the machine state of the physical machine management table 38 is changed to “maintenance stop requesting”, the load distribution means 45 will change the virtual servers 12a, 12b,. , 12m, 13a, 13b,..., 13m, 14a, 14b,..., 14m, 15a, 15b,.

FIG. 10 is a flowchart showing an operation at the time of the maintenance shift of the physical server according to the embodiment of the present invention.
10, the virtual machine abnormality monitoring means 37 in FIG. 3 reads the operation schedule of the physical servers 11a, 11b,..., 11n scheduled for maintenance from the allocation system management table 39 (step S81), and (planned maintenance start time− It is determined whether (terminal notification start time) <current time (step S82).

  If (maintenance start scheduled time−terminal notification start time) <current time, the item “machine state” in the physical machine management table 38 is changed to “maintenance stop request & message notification” (step S83). When the machine state of the physical machine management table 38 is changed to “maintenance stop request & message notification”, the application of each company stops the physical servers 11a, 11b..., 11n at the scheduled maintenance start time. The message is notified to the client terminal (step S84).

FIG. 11 is a flowchart showing an operation at the time of the maintenance shift of the physical server according to the embodiment of the present invention.
11, the virtual machine abnormality monitoring means 37 in FIG. 3 reads the operation schedules of the physical servers 11a, 11b,..., 11n scheduled for maintenance from the allocation system management table 39 (step S91), and scheduled maintenance start time <current It is determined whether it is time (step S92).

  If the scheduled maintenance start time is less than the current time, the virtual servers 12a, 12b,..., 12m, 13a, 13b running on the physical servers 11a, 11b,. .. 13m, 14a, 14b,..., 14m, 15a, 15b,..., 15m while determining whether there is no more request, the virtual servers 12a, 12b,. 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m are stopped (steps S93, S94).

  Then, all virtual servers 12a, 12b,..., 12m, 13a, 13b,..., 13m, 14a, 14b,. When 14m, 15a, 15b,..., 15m are stopped (step S95), stop requests for the physical servers 11a, 11b,..., 11n are written in the virtual machine switching schedule data 42 (step S96). ).

  Then, the virtual machine control instruction means 43 periodically refers to the virtual machine switching schedule data 42, and virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ... When a stop request for 14m, 15a, 15b, ..., 15m is found, the virtual servers 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ... ., 14m, 15a, 15b,..., 15m are instructed to stop (step S98).

It is a block diagram which shows schematic structure of the virtual server management system to which the load distribution apparatus which concerns on 1st Embodiment of this invention is applied. It is a figure which shows an example of the allocation method to the physical server of the system operated on the virtual server which concerns on one Embodiment of this invention. It is a block diagram which shows schematic structure of the virtual server management system with which the load distribution apparatus which concerns on 2nd Embodiment of this invention is applied. 4 is a flowchart illustrating a method for assigning a system operated on a virtual server using a genetic algorithm according to an embodiment of the present invention to a physical server. It is a flowchart which shows the method to allocate the load of a physical server with respect to the system operate | moved on the virtual server which concerns on one Embodiment of this invention. It is a flowchart which shows the method to allocate the load of a physical server with respect to the system operate | moved on a virtual server at the time of the overload of the physical server which concerns on one Embodiment of this invention. It is a flowchart which shows the method to allocate the load of a physical server with respect to the system operate | moved on a virtual server at the time of abnormality of the physical server or virtual server which concerns on one Embodiment of this invention. It is a figure which shows the method to allocate the load of a physical server with respect to the system operate | moved on a virtual server at the time of the maintenance transfer of the physical server which concerns on one Embodiment of this invention, or at the time of abnormality of a virtual server. It is a flowchart which shows the operation | movement at the time of the stop of new request reception at the time of the maintenance transfer of the physical server which concerns on one Embodiment of this invention. It is a flowchart which shows the operation | movement at the time of the terminal notification start at the time of the maintenance transfer of the physical server which concerns on one Embodiment of this invention. It is a flowchart which shows the operation | movement at the time of the maintenance transfer of the physical server which concerns on one Embodiment of this invention.

Explanation of symbols

1a, 1b, ..., 1n, 11a, 11b, ..., 11n Physical servers 2a, 2b, ..., 2m, 3a, 3b, ..., 3m, 4a, 4b, ..., 4m 5a, 5b, ..., 5m, 12a, 12b, ..., 12m, 13a, 13b, ..., 13m, 14a, 14b, ..., 14m, 15a, 15b, ..., 15m Virtual server 6 Load status collection means 7 Load allocation means 8 Load distribution means 9 Load distribution device 16 Virtual machine management manager 17 Operation status collection means 18 Operation result transmission means 19, 36 Operation result data 20 Operation result collection means 21 Virtual machine control means 23, 32 Transmission / reception control means 31 Load balancer 33 Operating status receiving means 34 Operating result receiving means 35 Virtual machine status table 37 Virtual machine abnormality monitoring means 38 Physical machine Thin management table 39 Allocation system management table 40 Virtual machine allocation scheduler 41 Hourly operation result data 42 Virtual machine switching schedule data 43 Virtual machine control instruction means 45 Load distribution means

Claims (7)

Load status collection means for collecting the load status of a system operating on a plurality of virtual servers from a plurality of physical servers;
Load allocation means for allocating a load applied to the physical server according to a genetic algorithm in which a load capacity of a system operated by the plurality of virtual servers is expressed by a gene sequence;
And a load distribution unit that distributes the load assigned by the load allocation unit to the physical server,
The load allocating means is
While setting a reference unit obtained by dividing the processing capacity of the processor of the physical server into a plurality of genes as one gene,
By assigning individual symbols to each of a plurality of systems operated by the plurality of virtual servers, and arranging the number of symbols based on a value obtained by dividing the load capacity of each system by the reference unit , Each of these multiple systems is represented by a gene consisting of multiple identical symbols,
By arranging the number of symbols of each system in accordance with the priority order of the plurality of systems, the entire plurality of systems are represented by a single gene sequence, and the genes of the initial solid selection width are expressed in accordance with the initial solid generation order. A random number is used to select a gene to generate a plurality of initial solids.When the genes with the earliest order in the initial solid selection width are selected by the number of the system, the selection target is selected in the initial solid selection width. Change to the next gene, until the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems, repeat the process after the process of generating the initial solid using the random number,
When the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems, the solid gene obtained by repeating the selection and crossover from the plurality of initial individuals according to the genetic algorithm a load balancer, characterized in sequential assignment Rukoto to the physical servers within the load capacity set in the physical server to be allocated object.   The load distribution apparatus according to claim 1, wherein the load allocation unit allocates the load so that the load applied to the physical server falls within a predetermined range of the processing capacity of the physical server.   The load allocating means reallocates the load applied to the physical server so that the load applied to the physical server does not exceed the upper limit value when the load applied to the currently operating physical server exceeds the upper limit value. The load distribution apparatus according to claim 1.   When there is a plan to stop a physical server on which the system is operated on the virtual server, the load allocation unit reallocates the load of the system operated on the physical server on which the system is scheduled to stop to another physical server. The load distribution apparatus according to claim 1. Multiple physical servers,
A plurality of virtual servers each started on the physical server;
A load distribution device that calculates a load allocated to each physical server according to a genetic algorithm in which a load capacity of a system operated by the virtual server is expressed by a gene sequence, and distributes the calculated load to the physical server; , equipped with a,
The load balancer is:
While setting a reference unit obtained by dividing the processing capacity of the processor of the physical server into a plurality of genes as one gene,
By assigning individual symbols to each of a plurality of systems operated by the plurality of virtual servers, and arranging the number of symbols based on a value obtained by dividing the load capacity of each system by the reference unit , Each of these multiple systems is represented by a gene consisting of multiple identical symbols,
By arranging the number of symbols of each system in accordance with the priority order of the plurality of systems, the entire plurality of systems are represented by a single gene sequence, and the genes of the initial solid selection width are expressed in accordance with the initial solid generation order. A random number is used to select a gene to generate a plurality of initial solids.When the genes with the earliest order in the initial solid selection width are selected by the number of the system, the selection target is selected in the initial solid selection width. Change to the next gene, until the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems, repeat the process after the process of generating the initial solid using the random number,
When the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems, the solid gene obtained by repeating the selection and crossover from the plurality of initial individuals according to the genetic algorithm the virtual server management system, characterized in sequential assignment Rukoto to the physical servers within the load capacity set in the physical server to be allocated object. Collecting the load status of a system running on multiple virtual servers from multiple physical servers;
Calculating a load allocated to each physical server according to a genetic algorithm in which a load capacity of a system operated by the virtual server is represented by a gene sequence;
Distributing the load allocated to each physical server to the physical server, and causing the computer to execute the step .
In the step of calculating the load allocated to each physical server,
While setting a reference unit obtained by dividing the processing capacity of the processor of the physical server into a plurality of genes as one gene,
By assigning individual symbols to each of a plurality of systems operated by the plurality of virtual servers, and arranging the number of symbols based on a value obtained by dividing the load capacity of each system by the reference unit , Each of these multiple systems is represented by a gene consisting of multiple identical symbols,
By arranging the number of symbols of each system in accordance with the priority order of the plurality of systems, the entire plurality of systems are represented by a single gene sequence, and the genes of the initial solid selection width are expressed in accordance with the initial solid generation order. A random number is used to select a gene to generate a plurality of initial solids.When the genes with the earliest order in the initial solid selection width are selected by the number of the system, the selection target is selected in the initial solid selection width. Change to the next gene, until the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems, repeat the process after the process of generating the initial solid using the random number,
When the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems, the solid gene obtained by repeating the selection and crossover from the plurality of initial individuals according to the genetic algorithm Are sequentially allocated to the physical servers within the range of the load capacity set for the physical servers to be allocated . Collecting the load status of a system running on multiple virtual servers from multiple physical servers;
Calculating a load allocated to each physical server according to a genetic algorithm in which a load capacity of a system operated by the virtual server is represented by a gene sequence;
Be one that is executed the steps of dispersing a load assigned to each of the physical servers on the physical server, to a computer,
In the step of calculating the load allocated to each physical server,
While setting a reference unit obtained by dividing the processing capacity of the processor of the physical server into a plurality of genes as one gene,
By assigning individual symbols to each of a plurality of systems operated by the plurality of virtual servers, and arranging the number of symbols based on a value obtained by dividing the load capacity of each system by the reference unit , Each of these multiple systems is represented by a gene consisting of multiple identical symbols,
By arranging the number of symbols of each system in accordance with the priority order of the plurality of systems, the entire plurality of systems are represented by a single gene sequence, and the genes of the initial solid selection width are expressed in accordance with the initial solid generation order. A random number is used to select a gene to generate a plurality of initial solids.When the genes with the earliest order in the initial solid selection width are selected by the number of the system, the selection target is selected in the initial solid selection width. Change to the next gene, until the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems, repeat the process after the process of generating the initial solid using the random number,
When the number of genes assigned to the initial solid matches the total number of genes of the plurality of systems, the solid gene obtained by repeating the selection and crossover from the plurality of initial individuals according to the genetic algorithm Are sequentially allocated to the physical servers within the range of the load capacity set for the physical servers to be allocated .

Images