JP5773142B2 - Computer system configuration pattern calculation method and configuration pattern calculation apparatus - Google Patents

Description

  The present invention relates to an improvement in a management method of a virtual computer system that provides a plurality of virtual computers by virtualizing computer resources of a physical computer.

  A virtualization system that realizes power saving and low cost by starting up a plurality of virtual servers in a physical server is widely used regardless of companies or individuals. Here, when constructing a virtualization system, points to be considered increase compared to the case of configuring an existing system. The first point to consider is which virtual server is placed on which physical server. The second point is how the virtual servers are configured, and whether the number of virtual servers is appropriate. Here, the configuration refers to a connection relationship between a virtual server, a virtual switch, a virtual router, and the like in the virtualization system. The third point is how the physical server group that is the base of the virtualization system is configured, and whether the number of physical servers is appropriate. This configuration refers to the connection relationship between a physical server, a physical switch, a physical router, and the like that are the basis of the virtualization system.

  Which virtual server is allocated to which physical server greatly affects the performance of the virtualized system. For example, it is beneficial to place a virtual server that performs processing with high CPU load on a physical server with high CPU performance, and to place a virtual server that performs processing with high CPU load on a physical server with low CPU performance It can be said that it is not useful. If the communication frequency between a certain virtual server A and a certain virtual server B is high, it is desirable that the physical servers A and B on which the virtual servers A and B are arranged are also arranged nearby.

  How virtual servers are configured to create a virtualized system and whether the number of virtual servers is appropriate also affect the performance of the virtualized system. For example, if a configuration is adopted in which communication to a certain virtual server is concentrated, it can be said that the virtual server becomes a bottleneck and the configuration needs to be reviewed. In addition, when the number of virtual servers is insufficient, the load on the entire system becomes excessive, and when there are too many virtual servers, several idle virtual servers appear. Furthermore, since the performance of the virtual server depends on the physical server to be arranged, it is difficult to accurately estimate how many virtual servers are required.

  The configuration of the physical server and the number of physical servers that are the basis of the virtualization system are also important. For example, when there are too many servers connected to the switch, the switch becomes a bottleneck, and the connected physical server cannot exhibit sufficient performance. In addition, when the number of physical servers is insufficient, the load on the entire system increases.

  As a method for constructing a virtualization system, in Patent Document 1, each virtual server at each time when each virtual server is operated on one of a plurality of physical servers from actually measured data indicating the performance of the virtual server every predetermined time. It is proposed to calculate a combination of a virtual server and a physical server that maximizes the total performance value and to relocate each virtual server according to the combination. This makes it possible to place the virtual server on the physical server with the best compatibility.

  Patent Document 2 proposes that each virtual server be placed on a physical server so that the virtual server group does not share a single point of failure, using information related to the configuration of the virtual server when building a virtualization system. Yes. As a result, it is possible to construct a virtualization system in consideration of the failure point.

JP 2005-115653 A JP 2009-217434 A

  However, in the virtualization system construction method according to Patent Document 1 and Patent Document 2, only the part of which physical server the virtual server is arranged is considered. If the method of Patent Document 1 is realized, a virtual server is most efficiently allocated to a physical server when there is a bottleneck in the configuration of the base physical server or when the number of virtual servers is insufficient. Even so, performance improvements are not expected.

  Similarly, when the method of Patent Document 2 is realized, the information regarding the configuration of the virtual server is used, but there is no function of changing the configuration, which is also between the physical servers and between the virtual servers. Even if a bottleneck has occurred, it cannot be resolved.

  Especially in an environment where multiple virtual servers operate on a large number of physical servers in a data center or the like, it is necessary to quickly set combinations and arrangements that optimize the load on the computers (physical servers and virtual servers) and the network load. is there. However, there has been a problem that even if the above conventional example is used, it is not possible to quickly obtain a combination or arrangement that optimizes the load on the computer or the network.

  Therefore, the present invention has been made in view of the above-described problems. When considering the arrangement of virtual servers and physical servers, the configuration of the entire virtualization system according to the user's needs, taking into account the physical configuration and logical configuration. The purpose is to quickly calculate.

The present invention provides a method in which a management computer having a processor and a memory calculates a configuration pattern of elements constituting a computer system, wherein the management computer includes configuration information including information on elements constituting the computer system. A first step of acquiring, a second step of acquiring load information including a load of an element constituting the computer system by the management computer, and a constraint condition for the configuration pattern of the computer system by the management computer. A third step of acquiring, a fourth step in which the management computer generates a plurality of configuration patterns based on configuration information of the computer system, and the management computer among the plurality of configuration patterns generated A fifth step of correcting a configuration pattern that does not satisfy the constraint condition, and the management computer Referring to information, and a sixth step of calculating a predetermined fitness for each of the constraint condition is satisfied constituting pattern and the modified configuration pattern, wherein the management computer, the fitness exceeds a predetermined threshold value If the configuration pattern is present, a seventh step of outputting the configuration pattern, the management computer, if the fitness is not present configuration pattern exceeds a predetermined threshold value, the fitness of a given An eighth step of applying a step of deleting a configuration pattern below a threshold, a step of replacing a configuration pattern with a portion of another configuration pattern, and a step of changing a portion of the configuration pattern. , returning to the fifth step after the eighth step, the first step, and physical components of the computer system, a logical structure Configuration information including raw information, and the second step acquires load information including a physical component load and a logical component load of the computer system, and The fourth step generates a configuration pattern of physical components and a configuration pattern of logical components of the computer system, and the fifth step includes the generated plurality of physical configurations. Among the patterns, the physical configuration pattern that does not satisfy the constraint condition is corrected, and among the plurality of generated logical configuration patterns, the logical configuration pattern that does not satisfy the constraint condition is corrected, The sixth step refers to the physical configuration information and the physical load information, and each of the physical configuration pattern satisfying the constraint condition and the modified physical configuration pattern A predetermined fitness is calculated for each of the logical configuration pattern and the logical load information, and the logical configuration pattern satisfying the constraint condition and the modified logical configuration pattern are respectively determined with reference to the logical configuration information and the logical load information. In the seventh step, there is a physical configuration pattern in which the fitness exceeds a predetermined threshold, and a logical configuration pattern in which the fitness exceeds a predetermined threshold. If so, the physical configuration pattern and the logical configuration pattern are output .

  Therefore, in the present invention, it is possible to calculate a configuration pattern of a computer system having a high degree of freedom that cannot be realized by the conventional technology by considering the connection relationship and the arrangement relationship of the components. In addition, if the user inputs needs to the management computer, the configuration of the entire computer system can be calculated quickly, eliminating the need for the user to study the configuration with a great deal of labor, and reducing the management cost. Is possible.

It is a block diagram which shows embodiment of this invention and shows the structure of a virtual machine system. 1 is a block diagram illustrating an example of a data center that operates a virtual machine system according to an embodiment of this invention. It is a block diagram which shows embodiment of this invention and shows an example of a calculation server. It is a block diagram which shows embodiment of this invention and shows an example of a client. It is a block diagram which shows embodiment of this invention and shows the connection relation of a virtual computer resource. It is a figure which shows embodiment of this invention and shows an example of the structure information which shows the connection relation of a virtual computer resource. It is a block diagram which shows embodiment of this invention and shows the connection relation of a physical computer resource. It is a figure which shows embodiment of this invention and shows an example of the structure information which shows the connection relation of a physical computer resource. It is a block diagram which shows embodiment of this invention and shows the correspondence of a virtual computer resource and a physical computer resource. FIG. 6 is configuration information showing an embodiment of the present invention and showing a correspondence relationship between virtual computer resources and physical computer resources. FIG. It is a figure which shows embodiment of this invention and shows an example of the load information of a virtual computer resource. It is a figure which shows embodiment of this invention and shows an example of the load information of a physical computer resource. It is a figure which shows embodiment of this invention and shows an example of the load information of a calculation server physical computer resource when calculating structure information. It is a sequence diagram which shows embodiment of this invention and shows the procedure which calculates structure information in time series. It is a flowchart which shows embodiment of this invention and shows an example of the process which calculates structure information. It is a flowchart which shows embodiment of this invention and shows the other example of the process which calculates structure information. It is a figure which shows embodiment of this invention and shows an example of the procedure which produces | generates the gene which shows the physical structure of a virtual computer system. It is a figure which shows embodiment of this invention and shows an example of the procedure which produces | generates the gene which shows the logical structure of a virtual computer system. It is a cluster table of genes which shows embodiment of this invention and shows the physical structure of a virtual computer system. It is a cluster table of genes which shows embodiment of this invention and shows the logical structure of a virtual computer system. It is a figure which shows embodiment of this invention and shows an example of a request / constraint reflection table. It is a figure which shows embodiment of this invention and shows an example of server information. It is a figure which shows embodiment of this invention and shows an example of the load information of a virtual server. It is a figure which shows embodiment of this invention and shows an example of the load information of a physical server. It is a figure which shows embodiment of this invention and shows an example of the basic load information of a virtual server. It is a figure which shows embodiment of this invention and shows an example of a lethal filter. It is a figure which shows embodiment of this invention and shows the procedure which calculates | requires the height of the load of the virtual server 403. FIG. It is a figure which shows embodiment of this invention and shows the procedure which calculates | requires change man-hour. 4 is a flowchart of cluster adjustment processing according to the embodiment of this invention. It is explanatory drawing of cluster operation in the case of showing the embodiment of this invention and creating a new cluster. It is explanatory drawing of cluster operation in the case of showing embodiment of this invention and increasing the number of genes to the existing cluster. It is a figure which shows embodiment of this invention and shows the procedure which calculates | requires the cost for every classification of a physical server. FIG. 5 is a sequence diagram illustrating a configuration change including a physical configuration change according to the embodiment of this invention. The embodiment of the present invention is shown, and a sequence diagram at the time of a configuration change not including a physical configuration change is shown.

  Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

  In this embodiment, an example using a genetic algorithm which is a kind of combination optimization algorithm as a method for calculating the configuration of a computer system (component connection relation and component arrangement) is shown below. Optimization algorithms and dynamic programming may be used. Although it is possible to solve without using a combinatorial optimization algorithm or dynamic programming, the computational complexity can be greatly reduced by using a combinatorial optimization algorithm or dynamic programming.

  FIG. 1 is a block diagram illustrating an embodiment of the present invention and an example of a data center that operates a virtual machine system.

  The computer system according to the first embodiment is a computer server that calculates the configuration of a virtual computer system when the client 101 constructs and manages the virtual computer system via the controller (virtualization management server) 201 in the data center 103. By using the management computer or configuration pattern calculation device 102, a virtual computer system corresponding to the needs of the administrator of the virtual computer system is calculated and constructed. Here, the configuration of the calculation server 102 is not particularly limited. For example, the client 101 and the calculation server 102 may be realized by a single unit having both functions, and the controller 201 and the calculation server 102 may be realized by a single unit having both functions.

  Here, as an example of the configuration calculated by the calculation server 102, an example of the configuration pattern of the components of the computer system including the connection relationship between the physical server 202 and the network device, the arrangement of the virtual server 403 operated on the physical server 202, and the like. Indicates.

  The client 101, the calculation server 102 and the router 203 of the data center 103 are connected via the network 104. The data center 103 includes a controller 201, a physical server 202, a router 203, and a switch 204, and the controller 201 is connected to all physical servers 202 via an internal LAN including the router 203 and switch 204. As will be described later, the controller 201 functions as a virtualization management server that manages the execution, stop, or movement of a virtual server running on the physical server 202.

  FIG. 2 is a block diagram showing the configuration of the data center 103. The controller 201 is connected to the router 203 in the data center 103. The physical server 202 is connected to the router 203 in the data center 103 via the switch 204. The router 203 in the data center 103 is connected to the network 104, and the controller 201 and the physical server 202 are indirectly connected to the network 104.

  The controller 201 includes a processor 210 that performs arithmetic processing, a memory 304 that stores data and programs, and a network I / F 301, and is connected to the router 203 via the network I / F 301. Further, the controller 201 stores a program and data in a storage device (not shown).

  A configuration / load information acquisition unit 303 and a configuration change instruction unit 302 are loaded into the memory 304 and executed by the processor 210. The configuration / load information acquisition unit 303 acquires configuration information 701 and load information 702 from a plurality of physical servers 202.

  The memory 304 includes the configuration information of the virtual servers 304 configured on the plurality of physical servers 202, the configuration information of each physical server 202, and the definition of the virtual computer system such as the type of the virtual server 403 and the type of the physical server 202. Information is stored as configuration information 701. Also, the load information of each virtual server 403 and each physical server 202 obtained by the configuration / load information acquisition unit 303 is stored in the memory 304 as load information 702.

  The physical server 202 includes a processor 220 that performs arithmetic processing, a memory 314 that stores data and programs, and a network I / F 401. The physical server 202 is connected from the network I / F 401 to the router 203 via the switch 204.

  The memory 314 is loaded with the hypervisor 402 and executed by the processor 220. The hypervisor 402 allocates computer resources of the physical server 202 to one or more virtual servers 403.

  When the configuration / load information acquisition request is received by the network I / F 301 from the calculation server 102 via the network 104, the controller 201 causes the configuration / load information acquisition unit 303 of all the physical servers 202 to pass through the network I / F 301. A command is transmitted to the hypervisor 402 to transmit the load information of each virtual server 403 and each physical server 202. Each hypervisor 402 transmits the detected load information to the controller 201, and the controller 201 stores the load information as load information 702 in the memory 304.

  Further, each hypervisor 402 transmits the detected configuration information to the controller 201, and the controller 201 stores the configuration information 701 in the memory 304.

  In addition to the load information 702 of each virtual server 403 and each physical server 202 stored in the memory 304, the controller 201 transmits the configuration information 701 on the memory 304 to the calculation server 102 via the network I / F 301.

  In addition, when a configuration change request is received from the calculation server 102 via the network 104 by the network I / F 301, the configuration change instruction unit 302 of the controller 201 uses the network I / F 301 to change the physical server 202 that requires the configuration change. A configuration change command is transmitted to the hypervisor 402. After a sufficient amount of time has passed for the change to be reflected, the configuration change instruction unit 302 correctly changes the configuration to the hypervisor 402 of the physical server 202 that has changed the configuration via the network I / F 301. Detect whether or not When the configuration change has been performed as requested, the configuration change instruction unit 302 notifies the calculation server 102 of the completion of the configuration change via the network 104 via the network I / F 301 and the router 203. On the other hand, if the configuration change has not been performed normally, the configuration change instruction unit 302 performs the configuration change procedure again until the request is reflected. Alternatively, a configuration change error notification is transmitted to the calculation server 102.

  The physical server 202 includes a host bus adapter (not shown) and accesses a storage device (not shown).

  FIG. 3 is a configuration diagram of the calculation server 102. The calculation server 102 includes a processor 230 that performs arithmetic processing, a memory 504 that stores data and programs, a storage device 231, and a network I / F 501, and is connected to the network 104 via the network I / F 501. A request acquisition unit 502, a configuration / load information acquisition unit 503, a configuration calculation unit 505, a configuration presentation unit 506, and a configuration change instruction unit 507 are loaded from the storage device 231 and executed by the processor 230 in the memory 504. The storage device 231 also functions as a storage medium for the request acquisition unit 502, the configuration / load information acquisition unit 503, the configuration calculation unit 505, the configuration presentation unit 506, and the configuration change instruction unit 507.

  The memory 504 stores configuration information 801 and load information 802 acquired by the configuration / load information acquisition unit 503 from the controller 201. The configuration information 801 includes configuration information of the virtual server 403 of the data center 103, configuration information of each physical server 202, information such as a virtual server type and a physical server type. Further, load information of each virtual server 403 and each physical server 202 is configured as load information 802.

  In addition, the request acquisition unit 502 stores the request received from the client 101 or the like, the constraint condition, or information related to the calculated configuration in the memory 504 as the request information 803. The calculation server 102 stores information necessary for calculating the configuration of the virtual machine system in the memory 504 as system definition information 804.

  Here, the system definition information 804 includes server information 1001, a request / constraint reflection table 1002, and a lethal filter 1003. Server information 1001 stores information related to the performance of the physical server 202. The request / constraint reflection table 1002 stores information on how the request information 803 is reflected when the configuration of the virtual machine system is calculated. The lethal filter 1003 stores information related to the operation of the genetic algorithm used in calculating the configuration of the virtual machine system. The system definition information 804 needs to be input before calculating the configuration of the virtual machine system, and may be input from the input device 232 or remote input from the client 101.

  When a request or restriction condition and a configuration change request are received by the network I / F 501 from the client 101 via the network 104, the request acquisition unit 502 stores the received request or restriction condition as request information 803 in the memory 504, A configuration / load information acquisition request is sent to the configuration / load information acquisition unit 503.

  Upon receiving the configuration / load information acquisition request, the configuration / load information acquisition unit 503 transmits the configuration / load information acquisition request to the controller 201 in the data center 103 via the network I / F 501 and the network 104.

  The configuration information of the virtual server in the virtual machine system, the configuration information of the physical server, and the load information of each virtual server and physical server are transferred from the controller 201 in the data center 103 via the network 104 to the network I / F 501 of the calculation server 102. Received. The configuration / load information acquisition unit 503 stores the received information as configuration information 801 and load information 802 in the memory 504 and transmits a configuration calculation request to the configuration calculation unit 505.

  Upon receiving the configuration calculation request, the configuration calculation unit 505 receives the configuration information 801, the load information 802, the request information 803, the system definition information 804, the physical cluster table 851, and the logical cluster table 852 in the memory 504. Is calculated. When calculating the configuration, the configuration calculation unit 505 stores the processing traffic 900, the basic load 910, the information 920 regarding the calculated configuration, the virtual server assumed load 930, the physical server assumed load 940, and the load determination result 950 on the memory 504. Hold on. The physical cluster table 851 and the logical cluster table 852 can be generated when the configuration is calculated.

  If the configuration of the physical server 202 needs to be changed, the configuration calculation unit 505 transmits the result to the configuration presentation unit 506. The configuration presentation unit 506 transmits the configuration change content to the client 101 via the network I / F 501 and the network 104. When the configuration of the physical server 202 needs to be changed, the configuration presentation unit 506 transmits the calculation result of the configuration of the virtual machine system to the configuration change instruction unit 507, and the configuration change instruction unit 507 transmits the network I / F 501. Then, a configuration change request is transmitted to the controller 201 in the data center 103 via the network 104.

  Thereafter, when the configuration change completion notification is received by the network I / F 501 from the controller 201 in the data center 103, the configuration presentation unit 506 transmits the configuration change content to the client 101 via the network I / F 501 and the network 104. . Alternatively, when a configuration change error notification is sent from the controller 201 in the data center 103, the configuration presentation unit 506 transmits the calculated configuration and change error content to the client 101 via the network I / F 501 and the network 104. To do.

  FIG. 4 is a block diagram showing the configuration of the client 101. The client 101 includes a processor 240 that performs arithmetic processing, a memory 607 that stores data and programs, an I / O 606 and a network I / F 601, and is connected to the network 104 via the network I / F 601. An input device 611 and an output device 612 are connected to the I / O 606. The input device 611 is configured with a keyboard, a mouse, and the like. The output device 612 includes a display or the like.

  The memory 607 stores a configuration acquisition unit 602, a configuration display unit 603, a request acquisition unit 604, and a request transmission unit 605, and is executed by the processor 240. Here, the configuration content of the virtual machine system received by the configuration acquisition unit 602 is stored in the memory 607 as configuration information 901. The memory 607 stores request / restriction conditions acquired by the request acquisition unit 604 as request information 802. A request (required by a user (or administrator) regarding the calculation or change of the configuration of the virtual machine system) and a constraint condition are received by the request acquisition unit 604 from the input device 611 via the I / O 606.

  Upon receiving the user request and the constraint condition, the request acquisition unit 604 stores the received information as the request information 902 in the memory 607 and notifies the request transmission unit 605 of the completion of the storage of the request / restriction condition. Upon receiving the request / constraint condition storage completion notification, the request transmission unit 605 transmits the request information 902 stored in the memory 607 to the calculation server 102 via the network I / F 601 and the network 104. When the configuration change content is received by the network I / F 601 from the calculation server 102 via the network 104, the configuration acquisition unit 602 stores the received configuration change content in the memory 607 as the configuration information 901, and the configuration display unit 603 Notify completion of storage of configuration changes. The configuration display unit 603 that has received the storage completion notification of the configuration change content outputs the content of the configuration information 901 to the output device 612 via the I / O 606.

  When the configuration and error contents of the virtual machine system are received by the network I / F 601, the configuration acquisition unit 602 stores the received configuration in the memory 607 as configuration information 901, and the storage of the configuration information in the configuration display unit 603 is completed. To be notified. Upon receiving the configuration information storage completion notification, the configuration display unit 603 outputs the content of the configuration information 901 and the error content to the output device 612 via the I / O 606.

  5A, 5B, 6A, 6B, 7A, and 7B show the contents on the configuration information 701 in the data center 103, the configuration information 801 in the calculation server 102, and the configuration information 901 in the client 101. FIG. The configuration information 701, 801, and 901 are the same information. The configuration information 701 includes configuration information 701A in FIG. 5B, configuration information 701B in FIG. 6B, and configuration information 701C in FIG. 7B. The configuration information 801 and 901 are the same, and are configured from configuration information 801A to 801C and 901A to 901C, respectively. These configuration information are collectively referred to as configuration information 701, 801, and 901.

  As shown in FIG. 7A, the configuration information 701, 801, and 901 in FIGS. 5B, 6B, and 7B are stored in the virtual servers (NODE0, NODE1, and NODE2) on the three physical servers (Machine0, 1, and 2) 202. The physical server (Machine0, Machine1) is connected to the switch (SW0), the physical server (Machine2) is connected to the switch (SW1), and the two switches (SW0, SW1) 204 are connected to the router (Router0) 203. An example is shown.

  FIG. 5A shows a connection relationship of virtual computer resources such as a virtual server and a virtual switch. For example, two virtual servers NODE0 and NODE1 are virtual switches (Virtual SW0) in the hypervisor 402 of a physical server (Machine0). And one virtual server NODE2 is connected to a virtual switch (Virtual SW1) in the hypervisor 402 of the physical server (Machine 1). Then, two virtual switches (Virtual SW0 and Virtual SW1) are connected to a virtual router (Virtual Router0) configured by a software router operating on the hypervisor 402. In the case of such a configuration, configuration information 701A (801A, 901A) indicating the connection destination of the virtual server is expressed as shown in FIG. 5B. In FIG. 5B, one entry is configured from a virtual computer resource name 711 and a connection destination name 712. A virtual computer resource in one physical computer may include at least a node and a virtual switch.

  FIG. 6A shows the relationship between the physical server 202 and the physical network connection. Two physical servers (Machine 0, 1) are connected to the switch (SW1) 204, and one physical server (Machine 2) is the switch (SW1). And two switches (SW0, 1) are connected to the router (Router0) 203. The type of the physical server (Machine 0, 1) indicates “A”, and the type of the physical server (Machine 2) indicates “B”.

  FIG. 6B shows configuration information 701B (801B, 901B) indicating the connection relation of physical computer resources. In FIG. 6B, one entry is configured from a physical computer resource name 721, a connection destination name 722, and a computer resource type 723. When the types of the physical servers 202 are the same, it indicates that the servers have the same specifications. The detailed information of the server type is expressed in a form as shown in FIG.

  FIG. 7A is a diagram showing a correspondence relationship between physical computer resources and virtual computer resources indicating which virtual server is allocated to which physical server. As shown in the figure, the virtual servers NODE0 and NODE1 are arranged on the physical server Machine0, and the virtual server NODE2 is arranged on the physical server Machine2. In the illustrated example, a virtual server is not generated in the physical server (Machine 2).

  FIG. 7B shows configuration information 701C (801C, 901C) indicating the correspondence between physical computer resources and virtual computer resources indicating which physical server 202 the virtual server 403 is located on. In FIG. 7B, one entry is configured from the name 731 of the virtual computer resource and the name 732 of the physical computer resource that provides the virtual computer resource.

  8 and 9 show the contents of the load information 702 in the data center 103 and the load information 802 in the calculation server 102. The load information 702 and the load information 802 are the same. The load information 702 and 802 includes virtual computer resource load information 702A and 802A shown in FIG. 8, and physical computer resource load information 702B and 802B shown in FIG.

  FIG. 8 shows virtual computer resource load information 702A and 802A indicating each virtual server 403 and its CPU, memory, I / O, and traffic load. The load information 702A and 802A includes a virtual server name 741 for storing the name of the virtual server 403, a CPU load 742 for storing the usage rate of the processor 220 allocated to the virtual server, and the use of the memory 314 allocated to the virtual server. A memory load 743 for storing a rate, an I / O load 744 for storing a bandwidth usage rate of a storage device (not shown) accessed by the virtual server, and a traffic load 745 for storing a network load accessed by the virtual server. One entry is configured. The I / O load 744 and the traffic load 745 use the usage rate of a virtual device (virtual network I / F, etc.) assigned to each virtual server 403 by the hypervisor 402. The usage rate can be appropriately used from an average value or a maximum value within a predetermined time.

  FIG. 9 shows physical computer resource load information 702B and 802B indicating the physical servers 202 and their respective CPUs, memories, I / Os, and traffic loads.

  The load information 702B and 802B store the physical server name 751 for storing the name of the physical server 202, the CPU load 752 for storing the usage rate of the processor 220 of the physical server, and the usage rate of the memory 314 of the physical server. A memory load 753, an I / O load 754 that stores a bandwidth usage rate of a storage device (not shown) accessed by the physical server, and a traffic load 755 that stores a usage rate of a network I / F 401 accessed by the physical server An entry is constructed from. Each usage rate can be appropriately used from an average value or a maximum value within a predetermined time. The traffic load 755 may also hold the total traffic amount handled by the hypervisor 402 as information.

  FIG. 10 shows the contents of request information 803 in the calculation server 102 and request information 902 in the client 101. The request information 803 and the request information 902 are the same.

  As shown in the figure, the request information 803 (902) includes request contents 812 for items 811 such as urgency, upper limit man-hours, physical server number upper limit information, physical configuration changeability, and configuration requests. Here, the urgency is an index indicating how much time the customer can spend calculating the configuration, and if the calculation of the configuration can take time, the urgency is low, and the configuration calculation result is immediately If necessary, set the urgency level high.

  The urgency level takes a value from 0 to 1, for example, and the urgency level increases as the urgency level approaches 1. As will be described later, the degree of urgency is related to an item called “fitness” in the genetic algorithm. Further, the urgency level is a trade-off with the accuracy of the configuration to be calculated, and if the urgency level is high, that is, if time is not required for the configuration calculation, the validity of the calculated configuration is slightly lowered.

  The upper limit man-hour (the upper limit of man-hours in the figure) is the man-hour (number of changes) allowed by the client 101 when changing the configuration of the virtual machine system. Here, one man-hour for changing the virtual server 403 corresponds to migration (migration) of the virtual server 403, change of the connection destination of the virtual server 403, termination of the virtual server 403, and addition of the virtual server 403. Further, one man-hour in the physical server 202 corresponds to a change in the connection destination of the physical server 202 and an addition of the physical server 202.

  The upper limit on the number of physical servers sets how many physical servers 202 can be used for each type (A to C). Since the number of virtual servers depends on the number of physical servers, input of the upper limit number is not required. Further, whether or not the physical configuration can be changed is an input indicating whether or not to calculate the configuration of the virtual machine system in consideration of the change of the physical configuration. Changes in the logical configuration and changes in the placement of the virtual server 403 can be changed by the controller 201 in the data center 103. However, when changing the physical configuration, a user (such as an administrator of the data center 103) needs to make a direct change. Therefore, the client 101 can set whether or not to allow a change in physical configuration.

  In addition, the request for the configuration can be any request as long as it can be reflected as an evaluation index when calculating the physical configuration, but for example, items such as emphasizing the machine cost or minimizing the change man-hours are calculated. If prepared in advance by the server, the client 101 can be selected according to the desire of the administrator or the like.

  FIG. 11 is a sequence diagram illustrating an example of a procedure at the time of requesting the configuration of the virtual machine system. The configuration request is a command for the client 101 to cause the calculation server 102 to calculate the configuration of the virtual machine system.

  First, in the client 101, the request acquisition unit 604 receives a user (or administrator) request (configuration request) and constraint conditions for acquiring a system configuration from the input device 611 via the I / O 606, and requests information 902 in the memory 607. (S1).

  Next, the request transmission unit 605 of the client 101 reads the request information 902 from the memory 607, and transmits a configuration calculation request including a configuration request and a constraint condition to the calculation server 102 via the network I / F 601 and the network 104 (S2). .

  The configuration calculation request including the configuration request and the constraint condition is received by the request acquisition unit 502 via the network I / F 501 in the calculation server 102. The request acquisition unit 502 stores the received configuration request / restriction condition as request information 803 in the memory 504 (S3), and sends a configuration / load information acquisition request to the configuration / load information acquisition unit 503. Upon receiving the configuration and load information acquisition request, the configuration / load information acquisition unit 503 transmits the configuration / load information acquisition request to the controller 201 in the data center 103 via the network I / F 501 and the network 104 (S4).

  When the configuration / load information acquisition unit 303 in the controller 201 receives the configuration / load information acquisition request, the configuration / load information acquisition unit 303 transmits the virtual / virtual information to each hypervisor 402 of all physical servers 202 via the network I / F 301. A transmission command of configuration information and load information of the server 403 and each physical server 202 is output (S5).

  Upon receiving the configuration information and load information transmission command, each hypervisor 402 acquires the configuration information of each virtual server 403 and physical server 202 and the load measurement result (load information) and responds to the controller 201 ( S6).

  The controller 201 stores the configuration information and the load measurement result received from the hypervisor 402 in the memory 304 as configuration information 701 and load information 702. In addition to the load information 702 stored in the memory 304, the configuration information 701 existing on the memory 304 is transmitted to the calculation server 102 via the network I / F 301 (S7).

  When the network I / F 501 receives the configuration information of the virtual server 403 in the virtual computer system, the configuration information of the physical server 202, the definition information of the virtual computer system, and the load information of each virtual server 403 and each physical server 202, The configuration / load information acquisition unit 503 stores the received information in the memory 504 as configuration information 801 and load information 802, and transmits a configuration calculation request to the configuration calculation unit 505. Upon receiving the configuration calculation request, the configuration calculation unit 505 calculates the configuration as described later based on the configuration information 801, the load information 802, and the request information 803 in the memory 504. Then, the calculation server 102 outputs the calculated configuration information to the client 101 (S8).

  The processing after step S8 is the sequence diagram of FIG. 28 or FIG. 29 depending on whether or not the physical configuration is changed. When the physical configuration is changed, the processing is performed by the sequence diagram of FIG. If there is no change, the logical configuration is changed in the sequence diagram of FIG.

  FIG. 12 is a flowchart illustrating an example of a configuration calculation process using a genetic algorithm performed by the configuration calculation unit 505.

  A genetic algorithm is a learning algorithm that mimics the genetic mechanism of an organism, and a solution is derived by genetic operations such as selection, crossover, and mutation. As a method for calculating configuration information of a virtual computer system using a genetic algorithm, first, a plurality of initial populations as solution candidates are generated, and it is evaluated whether each individual (configuration) satisfies the request information 803.

  Next, as a result of evaluating the initial population, if there is an individual that satisfies the request information 803, the genetic algorithm is terminated. On the other hand, if there is no individual satisfying the request information 803 as a result of evaluating the initial population, genetic operations such as selection, crossover, and mutation are performed, and each individual is evaluated again. The configuration calculation unit 505 performs the above-described cycle for each of the physical configuration pattern and the logical configuration pattern, and repeats until an individual (configuration pattern) that satisfies the request information 803 appears.

  In the genetic algorithm, each individual is represented in the form of a gene. The gene expression method is not particularly limited as long as information necessary for gene manipulation can be expressed, but can be expressed as a gene representing the physical configuration of the virtual computer system, for example, by a numerical sequence as shown in FIG.

  FIG. 14 is a diagram illustrating an example of a procedure for creating a gene indicating a physical configuration pattern of a virtual computer system.

  As a method of creating a gene sequence representing the physical configuration of a computer system, first, the physical configuration of the virtual computer system is input (S31), and the connection relationship of each device (physical computer resource) is formed into a table and connected. In this case, its own type (server type) number is set, and if it is not connected, 0 is set (S32).

  In the example of FIG. 14, the physical server Sv1 is connected to the switch S1, and the physical server Sv2 is connected to the switch S2. The switches S1 and S2 are connected to the router R, and input = I is connected to the router R.

  In step S32, physical servers Sv1, Sv2, switches S1, S2, router R, and tables arranged in the row direction and the column direction in the order of input = I are formed. The type numbers are physical server = “1”, switch = “2”, and router = 3. First, in the row of the physical server Sv1, since the physical server Sv1 is connected to the switch S1, the type number “1” of the physical server is set in the column of the switch S1, and “0” is set in the other columns. Set. As a result, the row of the physical server Sv1 becomes “001000”. Similarly, since the physical server Sv2 is connected to the switch S2, the type number “1” of the physical server is set in the column of the switch S2, and “0” is set in the other columns. As a result, the row of the physical server Sv2 becomes “000100”. Similarly, since the switch S1 is connected to the physical server Sv1 and the router R, the physical server type number = “1” is set in the column of the physical server Sv1, and the switch type number is set in the column of the router R. = Set “2”. As a result, the row of the switch S1 becomes “100020”. The switch S2, the router R, and the input I are set similarly. Thereby, the configuration pattern of the computer system is expressed by a number sequence indicating a combination of components, and this number sequence can be handled as a gene pattern corresponding to the configuration pattern.

  Next, in step S33, the contents of the rows and columns of the table are symmetrical to each other, so that information necessary to express the configuration is half of the table (upper right or lower left). In the example shown in the figure, other information is deleted using the information in the upper right half.

  Furthermore, since it is not necessary to express that the devices themselves are connected, it is possible to omit another line of information from the half of the table (S34). That is, a column with the same row and column device is deleted.

  Next, “1. Physical servers are not connected to each other, 2. Switches are not connected to each other, 3. Physical servers are connected to physical routers via switches (physical servers are not directly connected to routers), 4, Input If the restriction that “only the router is connected to” is added, a portion that does not need to be expressed in the table can be further cut (S35).

  Only the necessary information is extracted from the table in step S35, and values arranged in order from the upper left column to the lower right column are gene sequences (S36). In the illustrated example, “1001223” is obtained.

  By the above procedure, the gene sequence takes a value of 0 or more, and a value of 0 indicates that it is not connected to the corresponding device, and a value of 1 or more indicates that it is connected to the corresponding device. In addition, the type of the device can be known by the size of a number having a value of 1 or more. Here, since the constraint applied in step S35 described above depends on the target virtual system, when applied to a system with a high degree of freedom, the expression in step S34 is not expressed in the gene expression without reducing the expression due to the constraint. And it is sufficient.

  However, if there is information that does not need to be expressed for the target virtual computer system, the length of the gene should be shortened. This also leads to a reduction in calculation amount in the calculation server 102.

  Further, the gene representing the logical configuration can be expressed by a numerical sequence as shown in FIG. 15, for example. FIG. 15 is a diagram illustrating an example of a procedure for creating a gene indicating the physical configuration of the virtual computer system.

  As a method of creating a gene sequence that represents the logical configuration of a computer system, first, the logical configuration of the virtual computer system is input (S41), and the connection relationship of each device (logical computer resource) is formed into a table and connected. If it is, the device number corresponding to its own device type is set, and if it is not connected, 0 is set (S42).

  In the example of FIG. 15, the virtual servers NODE1 (N1) and NODE2 (N2) of Machine1 are connected to the virtual switch S1 of Machine3, and the virtual servers NODE3 (N3) and NODE4 (N4) of Machine2 are connected to the virtual switch S2 of Machine3. Connected. The virtual switches S1 and S2 are each connected to the virtual router R of Machine 4, and the input I is connected to the virtual router R.

  In step S42, tables arranged in the row direction and the column direction in the order of virtual servers NODE1 to NODE4, virtual switches S1 and S2, virtual router R, and input I are formed. The device numbers are Machine1 = “1”, Machine2 = “2”, Machine3 = 3, and Machine4 = 4. First, in the row of the virtual server N1, since the virtual server N1 is connected to the virtual switch S1, the device number of the virtual server = “1” is set in the column of the switch S1, and “0” is set in the other columns. Set. As a result, the row of the virtual server N1 becomes “00001000”. Similarly, since the virtual server N3 is connected to the virtual switch S2, the device number of the virtual server N3 = “2” is set in the column of the virtual switch S2, and “0” is set in the other columns. As a result, the row of the virtual server N3 becomes “00000100”. Similarly, since the virtual switch S1 is connected to the virtual server N1 and the virtual router R, the device number of the virtual server N1 = “1” is set in the column of the virtual server N1, and the column of the virtual router R is set in the column of the virtual router R. The device number of the virtual switch = “3” is set. As a result, the row of the virtual switch S1 becomes “11000030”. The virtual switch S2, the virtual router R, and the input I are set similarly. As a result, the logical configuration pattern of the virtual computer system is expressed by a sequence of numbers indicating combinations of virtual components, and this sequence can be handled as a gene pattern corresponding to the configuration pattern.

  Next, in step S43, the contents of the row and column of the table are symmetrical with each other, so that the information necessary for expressing the configuration is half of the table (upper right or lower left). In the example shown in the figure, other information is deleted using the information in the upper right half.

  Furthermore, since it is not necessary to express that the device itself is connected, it is possible to omit another line of information from the half of the table (S44). That is, a column having the same row and column element is deleted.

  Next, “1. The virtual servers are not connected to each other, 2. The virtual switches are not connected to each other, 3. The virtual server is connected to the virtual router via the virtual switch (the virtual server is not directly connected to the virtual router), If the restriction that “4. Only a virtual router is connected to Input” is added, it is possible to further cut a portion that does not need to be expressed in the table (S45).

  Only the necessary information is extracted from the table in step S45, and the values sequentially arranged from the upper left column to the lower right column become the gene sequence (S46). In the illustrated example, “1010102202334” is obtained.

  With the above procedure, the gene arrangement also takes a value of 0 or more for the logical configuration, and a value of 0 indicates that it is not connected to the corresponding element, and a value of 1 or more indicates that it is connected to the corresponding element. In addition, the type of the device can be known by the size of a number having a value of 1 or more. Here, since the constraint applied in step S35 described above depends on the target virtual system, the expression in step S44 is not expressed in the table in step S44 without reducing the expression due to the constraint when applied to a system with a high degree of freedom. And it is sufficient.

  However, if there is information that does not need to be expressed for the target virtual computer system, the length of the gene should be shortened. This also leads to a reduction in calculation amount in the calculation server 102.

  Information may be collected into one gene that simultaneously represents the physical configuration and the logical configuration, but depending on the request information 803, there is a case where the change of the physical configuration is prohibited. In that case, by dividing the gene into a physical configuration and a logical configuration, it is only necessary to manipulate the gene representing the logical configuration, and thus the amount of calculation can be reduced. In addition, the complexity of the gene can be prevented.

  Here, in the gene expression of FIG. 14 and FIG. 15, the number of servers for each server type or device type cannot be represented, and the length of the gene varies depending on the number of servers, so that gene manipulation is difficult. Therefore, in the present invention, genes having the same number for each server type or device type are expressed as a cluster, and gene manipulation is performed in units of clusters. In other words, the cluster is a set of configuration patterns in which the lengths of the sequences indicating the genes are equal.

  FIG. 16 and FIG. 17 represent genes representing a physical configuration and clusters of genes representing a logical configuration. FIG. 16 shows a gene cluster table representing a physical configuration. FIG. 17 shows a gene cluster table representing a logical structure.

  Each cluster table is generated by the configuration calculation unit 505 based on the input configuration (physical configuration and logical configuration), and the number of each cluster server type or device type is the additional machine type in FIG. 16 and FIG. Can be calculated based on the additional VM type, and the number of genes and gene information held by each cluster. The cluster is stored in the memory 504 in the form as shown in FIGS. 16 and 17, and is rewritten as needed during application of the genetic algorithm. Since all the gene lengths are the same in the cluster, gene manipulation can be performed freely.

  In FIG. 16, a physical cluster table 851 includes a cluster number 8511 for storing a cluster identifier, an additional machine type 8512 for storing the type of a physical device to be added, and a gene number 8513 for storing the number of genes in the table. One entry is made up of a gene 8514 for storing a gene indicating a physical configuration.

  In the illustrated example, the cluster number 8511 = 0 stores a gene 8514 indicating a configuration example of four physical servers 202. The cluster number 8511 = 1 stores a gene 8514 when a physical server 202 of type A is added. Cluster number 8511 = 2 stores a gene 8514 when a physical server 202 of type A is added.

  In FIG. 17, the logical cluster table 852 includes a cluster number 8521 for storing the identifier of the cluster, an additional VM type 8522 for storing the type of the element to be added, the number of genes 8523 for storing the number of genes in the table, One entry is configured from a gene 8524 that stores a gene indicating a physical configuration.

  In the illustrated example, a cluster number 8521 = 0 stores a gene indicating a configuration example of six virtual servers 403. The cluster number 8511 = 1 stores a gene 8524 when one virtual server 403 is added. Cluster number 8521 = 2 stores gene 8524 when two virtual servers are added.

<Outline of processing>
FIG. 12 is a flowchart illustrating an example of a configuration calculation process performed by the calculation server 102 in consideration of the physical configuration of the virtual computer system, the logical configuration, and the placement of the virtual server.

  The request information and the configuration / load information collected by the request acquisition unit 502 and the configuration / load information acquisition unit 503 via the network I / F 501 are sent to the configuration calculation unit 505 via the memory 504, and the configuration calculation unit 505 is the client. When a command for calculating the configuration is received from 101, processing is performed according to the flowchart of FIG. This flowchart includes a loop (S12, S13, S21 to S25) for calculating a physical configuration pattern, and a loop portion (S14 to S20) for calculating a logical configuration pattern and a virtual server arrangement.

  As an example of the process of applying the genetic algorithm, an example in which the genetic algorithm for calculating the logical configuration and the server arrangement is nested in the genetic algorithm for calculating the physical configuration is shown.

  When the calculation server 102 executes the configuration calculation process, the configuration calculation unit 505 refers to the request information 803 and reflects it in the request / constraint reflection table 1002 in the system definition information 804. The request / constraint reflection table 1002 will be described later. The configuration calculation unit 505 of the calculation server 102 receives information 920 related to the configuration to be calculated in addition to the request information 803 from the client 101 before starting the configuration calculation processing. The configuration calculation unit 505 acquires the urgency a of the received request / constraint reflection table 1002, and obtains a fitness threshold as shown in FIG. The degree of urgency a is a parameter that determines the level of calculation accuracy (the length of calculation time), and is expressed as a value between 0 and 1.

  Next, the configuration calculation unit 505 estimates the load of each virtual server 403 in step S11 of FIG. For this load estimation, as shown in FIG. 20C described later, the load of the virtual server 202 per unit traffic is obtained as a basic load 910. Then, the configuration calculation unit 505 calculates the assumed load 930 of the virtual server 403 when processing the acquired traffic volume 923 and the assumed load 940 of the physical server 202 that provides each virtual server 403. Then, the configuration calculation unit 505 compares the assumed load with a predetermined default value and calculates a load determination result 950 for each of the virtual server 403 and the physical server 202. The processing traffic volume 923 is the traffic volume measured when the load information is acquired.

  Next, in step S12, the configuration calculation unit 505 generates an initial pattern of the physical configuration with a gene pattern for each cluster. As a method of generating a physical pattern, for example, a physical configuration pattern for each cluster can be generated by generating a random number. Then, the generated physical configuration pattern is converted into a gene by the above-described processing.

  Next, in step S13 (correction unit), the configuration calculation unit 505 determines that if the physical configuration gene generated in step S12 violates the constraint condition of the system definition information 804, the physical configuration gene is the constraint condition. Modify to meet This modification shows an example in which the lethal filter 1003 of the system definition information 804 is used as a constraint condition. As will be described later, the lethal filter 1003 stores conditions for determining genes that violate preset computer system definitions and contents for correcting genes that violate computer system definitions. The configuration calculation unit 505 determines whether or not each of the generated genes of the physical configuration matches the definition, and corrects the gene according to the correction contents if there is a physical configuration of the gene that violates the definition.

  Next, in step S14, the configuration calculation unit 505 generates an initial pattern of the logical configuration as a gene pattern for each cluster. As a method of generating a logical configuration, for example, a logical configuration pattern for each cluster can be generated by generating a random number. Then, the generated logical configuration pattern is converted into a gene by the above-described processing.

  Next, in step S15 (correction unit), the configuration calculation unit 505 determines that the gene satisfies the constraint condition if the gene of the logical configuration generated in step S14 violates the constraint condition of the system definition information 804. To correct. This modification shows an example in which the lethal filter 1003 of the system definition information 804 is used as a constraint condition as in the physical configuration. The configuration calculation unit 505 determines whether or not each of the generated genes of the logical configuration matches the definition, and corrects the gene according to the correction content if there is a gene that violates the definition.

  In step S16 (a fitness calculation unit), the configuration calculation unit 505 calculates the fitness for each logical configuration of each cluster from equation (1) described later. Then, the configuration calculation unit 505 compares the fitness of each logical configuration with a predetermined threshold value, and determines whether there is a logical configuration with a fitness level exceeding the threshold value. If there is a logical configuration having a fitness level that exceeds the threshold value in any cluster, the logical configuration is taken as a solution and the process proceeds to step S21 (a fitness level calculation unit). On the other hand, if there is no fitness exceeding the threshold, the process proceeds to step S17.

  If there is no fitness level equal to or greater than the preset threshold value, the configuration calculation unit 505 operates the genes of each logical configuration in steps S17 to S19 (configuration change unit). Here, as the threshold value for determining the fitness level, the threshold value for the fitness level set in the request / constraint reflection table 1002 in FIG. 18 is used.

  First, in step S <b> 17 of FIG. 12, the configuration calculation unit 505 performs a gene operation called selection on a logically configured gene. The selection process executed by the configuration calculation unit 505 is an operation for deleting genes having a logical configuration with a fitness less than a predetermined value and leaving a large number of individuals (configuration) with a high fitness (a predetermined value or more) in the next generation. It is. As an example of the selection process, a known or well-known method such as roulette selection, ranking selection, tournament selection, elite selection, or the like can be used. Here, as the gene, when the expressions as shown in FIGS. 14 and 15 are used, and as shown in FIGS. 16 and 17, the genes are handled as a group as a cluster.

  In step S <b> 18, the configuration calculation unit 505 adjusts each of the gene groups having the logical configuration of the plurality of clusters. The adjustment of each cluster executed by the configuration calculation unit 505 will be described later with reference to FIG.

  In step S <b> 19, the configuration calculation unit 505 performs crossover for the gene groups having the logical configuration of the selected clusters. Crossover is a process of taking two genes out of the genes selected by the selection and multiplying the genes. As the crossover method, a known or publicly known method such as one-point crossover, two-point crossover, multipoint crossover, or uniform crossover can be used.

  However, when the gene expression shown in FIGS. 8 and 9 is used, genes having different numbers of virtual servers and different physical servers have different gene lengths. For this reason, at the time of crossover, it is possible to crossover between genes in the cluster and crossover genes of the same server type.

  In step S <b> 20, the configuration calculation unit 505 performs a mutation that is the last step of genetic manipulation for the gene groups of the plurality of clusters that have undergone crossover. The genetic manipulation of mutation is an operation of changing a part of information of a bit string of gene expression with a low probability set in advance for an individual (gene) that has been genetically manipulated until crossover. Thereby, it can prevent that the individual (gene) to calculate falls into a local optimal solution.

  After completing all the genetic manipulation steps in steps S17 to S20, the configuration calculation unit 505 returns to the process of step S15. In step S15, the configuration calculation unit 505 causes the gene expression to satisfy the constraint condition if the gene expression of the logical configuration changed by the crossover or mutation violates the constraint conditions in the request information 803. Correction is made by the lethal filter 1003. After these processes, the fitness of each component is determined again (step S16 in FIG. 12).

  Next, when there is a logical configuration whose fitness exceeds the threshold value in the determination in step S16, the process proceeds to step S21, and the configuration calculation unit 505 calculates the fitness for each of the physical configurations of the plurality of clusters. The calculation of the fitness of the physical configuration will be described later. The configuration calculation unit 505 compares the fitness of physical configurations of a plurality of clusters with a predetermined threshold value, and determines whether there is a physical configuration having a fitness level exceeding the threshold value. Here, as the threshold value for determining the fitness level, the threshold value for the fitness level set in the request / constraint reflection table 1002 in FIG. 18 is used as in the determination in step S16.

If any cluster has a physical configuration having a fitness level that exceeds the threshold, the configuration calculation unit 505 sets the physical configuration as a solution and proceeds to step S26. On the other hand, if there is no fitness exceeding the threshold, the process proceeds to step S22. In step S26 , the configuration calculation unit 505 outputs the logical configuration and physical configuration of the fitness exceeding the threshold value, and the process ends.

  If there is no fitness level equal to or greater than a preset threshold value, the configuration calculation unit 505 operates the genes of each logical configuration in steps S22 to S25 (configuration change unit).

  First, in step S22, the configuration calculation unit 505 performs a genetic operation called selection on a gene having a physical configuration. The selection process executed by the configuration calculation unit 505 is an operation for deleting a gene having a physical configuration having a fitness less than a predetermined value and leaving a large number of individuals (configuration) having a high fitness (a predetermined value or more) in the next generation. It is. As an example of the selection process, a known or well-known method such as roulette selection, ranking selection, tournament selection, elite selection, or the like can be used. Here, the genes are the same as in the logical configuration described above, and when the expressions as shown in FIGS. 14 and 15 are used, the genes are handled as a group as a cluster as shown in FIGS. 16 and 17.

  In step S <b> 23, the configuration calculation unit 505 adjusts each of the gene groups having the physical configuration of the plurality of clusters. The adjustment of each cluster executed by the configuration calculation unit 505 will be described later with reference to FIG.

  In step S24, the configuration calculation unit 505 performs crossover for the gene group having the physical configuration of the selected clusters. Crossover is a process of taking two genes out of the genes selected by the selection and multiplying the genes. As the crossover method, a known or publicly known method such as one-point crossover, two-point crossover, multipoint crossover, or uniform crossover can be used.

  However, when the gene expression shown in FIGS. 8 and 9 is used, genes having different numbers of virtual servers and different physical servers have different gene lengths. For this reason, at the time of crossover, it is possible to crossover between genes in the cluster and crossover genes of the same server type.

  In step S <b> 25, the configuration calculation unit 505 performs a mutation that is the last step of genetic manipulation for the gene group having the physical configuration of the plurality of clusters that have undergone crossover. The genetic manipulation of mutation is an operation of changing a part of information of a bit string of gene expression with a low probability set in advance for an individual (gene) that has been genetically manipulated until crossover. Thereby, it can prevent that the individual (gene) to calculate falls into a local optimal solution.

  After completing all the genetic manipulation steps in steps S22 to S25, the configuration calculation unit 505 returns to the process of step S13. In step S <b> 13, the configuration calculation unit 505 determines that the gene expression satisfies the constraint condition when the gene expression of the physical configuration changed by crossover or mutation violates the constraint conditions in the request information 803. Correction is made by the lethal filter 1003. After passing through these processes, the process after step S21 is repeated through steps S14-S16 again.

  Through the above processing, the configuration calculation unit 505 outputs a combination of a logical configuration and a physical configuration whose fitness exceeds a threshold value.

Incidentally, as the threshold (1-a ^4/2) of the fitness of the request / constraints reflect table 1002 the threshold in the above example, the amount of urgency a (except 0 ≦ a ≦ 1) is high becomes if genetic manipulation increased If the urgency a is low, the amount of genetic manipulation is reduced and the calculation time is shortened. The fitness threshold may be set separately from the urgency a.

  As described above, when the fitness of each configuration is calculated and there is a configuration exceeding the fitness threshold, the configuration is output as a solution (step S26 in FIG. 12), and the process is terminated.

  If there is no configuration that has a fitness level that exceeds the threshold value, after performing processing such as selection, crossover, and mutation for individuals with the physical configuration in the same way as the genetic operation at the time of logical configuration / server allocation calculation (Figure 12 Step S22-25), the process proceeds to the logical configuration / server arrangement calculation process again. However, unlike the case of genetic manipulation of the logical configuration, the percentage of the total number of congested physical servers in all genes exceeding the predetermined number during the cluster manipulation is obtained during the genetic manipulation of the physical configuration. As a result, it is possible to search for the number of servers according to the congestion of physical servers.

<Request / constraint reflection table>
FIG. 18 shows an example of the contents of the request / constraint reflection table 1002. The request / constraint reflection table 1002 stores information used in the processing of FIG.

  The request / restriction reflection table 1002 includes request / restriction items 10021 such as urgency, upper limit man-hours, physical server number upper limit information, change of physical configuration, request regarding configuration set in the request information 803 (902), and these requests. Each request content 10022 with respect to the constraint item and information 10025 to be reflected in the genetic algorithm (GA: Genetic Algorithm in the figure) among the request content 10022 are stored. It is assumed that the information 10027 to be reflected on the GA is set in advance from the client 101. As information 10025 to be reflected in the GA, a threshold value of fitness, a weighting coefficient, and the like are set.

  The request / constraint reflection table 1002 shows a method of reflecting the request contents expressed by the request information 803 in variables, procedures, etc. at the time of configuration calculation. By changing the configuration calculation procedure according to the request / constraint reflection table 1002, a request from the user operating the client 101 can be utilized for the configuration calculation.

<Measurement of load>
In the process of FIG. 12, first, the configuration calculation unit 505 of the calculation server 102 estimates the load when processing the unit traffic (1000 traffic) for each virtual server 403 (step S11 in FIG. 12). The configuration calculation unit 505 reads the load information shown in FIGS. 8 and 9 and calculates the load applied to each virtual server 403 when processing the unit traffic (1000 traffic).

  Regarding the load of the virtual server 403, the configuration calculation unit 505 acquires the load information of the virtual server 403 from FIG. 8, acquires the configuration information 701A of FIG. 5B, and generates the load information of the virtual server shown in FIG. 20A. 20A includes virtual server 741 to traffic load 745 of load information 802A shown in FIG. 8, connection destination 712 of configuration information 801A of FIG. 5B, and physical server 732 of FIG. Further, the processing traffic amount 900 is combined. Here, the processing traffic amount is a value obtained by the calculation server 102 from the virtual server 403 on the traffic amount processed for each virtual server 703. Note that the load information of the virtual server in FIG. 20A may be held on the memory 504 of the calculation server 102.

  Moreover, the load of the physical server 202 acquires the load information of the physical server 202 from FIG. 9, acquires the configuration information 701B of FIG. 6B, and generates the load information of the physical server shown in FIG. 20B.

  20B is obtained by combining the physical server 751 to the traffic load 755 of the load information 802B shown in FIG. 9 with the connection destination 722 of the configuration information 801B of FIG. 6B. 20B may be held on the memory 504 of the calculation server 102.

  The configuration calculation unit 505 then estimates the load per unit traffic (1000) shown in FIG. 20C from the server information shown in FIG. 19, the virtual server load information shown in FIG. 20A, and the physical server load information shown in FIG. 20B. Is calculated and stored in the table of the basic load 910. The basic load 910 includes a virtual server 911 that stores the name of the virtual server 403, a CPU load 912 that stores the usage rate of the processor 220 assigned to the virtual server for processing the unit traffic, and a unit traffic. A memory load 913 for storing the usage rate of the memory 314 allocated to the virtual server, and an I / O load for storing the bandwidth usage rate of a storage device (not shown) accessed by the virtual server to process unit traffic One entry is composed of the traffic load 915 that stores the network load accessed by the virtual server in order to process the unit traffic.

  A method is generally known in which the estimated load value per unit traffic (1000) is obtained by dividing each currently obtained load 912-915 by the current amount of processing traffic.

  Note that the server information 1001 in the system definition information 804 may be used for the load estimation.

  FIG. 19 shows the contents of the server information 1001. The server information 1001 stores a physical server 10011 that stores a server type, a gene number 10012 that stores a number to be set in gene information, a CPU performance 10013 that stores information on processor performance, and information on memory performance. Memory performance 10014, I / O performance 10015 for storing information related to performance of I / O access to a storage device, etc., traffic performance 10016 for storing information related to performance of access to network etc., and executable virtual One entry includes a virtual server upper limit 10017 for storing the maximum value of the server and 10018 for storing a value related to the cost of the physical server 202.

  Here, the information 10013 to 10016 related to the performance and the information 10018 related to the cost are set in advance with relative values with a physical server that is a preset reference. Other values are values preset by an administrator or the like.

  The server information 1001 is preset with numerical values (ratio) indicating superiority or inferiority of performance with respect to the reference physical server, the number of executable virtual servers, the cost ratio with respect to the reference physical server, numbers used in gene expression, and the like. It is a table.

  An administrator or the like may appropriately set a reference server for comparing the superiority and inferiority of performance and cost as long as the performance and cost of each server can be expressed relatively from the reference server. For estimating the basic load 910, a known or well-known method can be adopted as appropriate.

  However, as shown in FIGS. 20A to 20C, estimation is performed using information such as configuration information 801, load information 802, and server information 1001 in system definition information 804 stored in the memory 504 in the calculation server 102 as input, As an output, it is required to output a load amount per unit traffic (1000 traffic) when starting with a reference server (each handling 1.0) for each virtual server.

<Fatality filter>
FIG. 21 shows an example of the lethal filter 1003.

  The lethal filter 1003 includes a set of a determination part 10031 for storing a condition for determining a gene that violates the system definition and a correction method 10032 for storing contents for correcting a gene that violates the system definition. Here, only the determination part and its correction procedure are described, but a filter may be described by a program. By defining outside as the lethal filter 1003, even when the system definition changes, the type of lethal gene can be changed only by operating this filter portion.

<Calculation of fitness of logical configuration>
The calculation of the fitness of the logical configuration performed in step S16 in FIG. 12 is performed by the following equation, for example.

  Here, “good_chemistry_virtual_machine” in the above equation (1) represents the number of virtual servers 403 that are compatible with the physical server 202. “Bad_chemistry_virtual_machine” represents the number of virtual servers 403 that are not compatible with the physical server 202.

  Here, in the present embodiment, the compatibility between the physical server 202 and the virtual server 403 means that the portion that is most loaded by the processing of the virtual server 403 (for example, CPU, memory, IO, NW, etc.) and the physical server 202 are good at it. It is defined that the compatibility is good when the virtual server 403 has the same processing. The compatibility between the physical server 202 and the virtual server 403 is obtained by referring to the basic load information of the virtual server 403 per unit traffic and the server information 1001 in the system definition information 804 described above.

  For example, in FIG. 20C, the item with the largest load on Node 0 is the traffic load. Here, referring to the traffic performance 10016 of the server information 1001, it can be understood that the server strong against the traffic load is the physical server B. Therefore, in the load situation shown in FIG. 20C, when Node 0 is arranged in the physical server B, it is determined that the compatibility is good.

  Next, “conquestion_virtual_machine” in the above formula (1) represents the number of virtual servers 403 having a high load, and “free_virtual_machine” represents the number of virtual servers 403 having a low load. The high load of the virtual server can be obtained from the basic load per unit traffic, the arrangement of the virtual server having the calculated configuration, the rough estimate of the processing traffic amount 900, and the server information 1001.

  FIG. 22 shows a procedure for obtaining the load level of the virtual server 403. First, the configuration calculation unit 505 calculates and calculates a basic load 910 (1000 traffic units in FIG. 22) per unit traffic when starting up with a reference server (each performance (ratio) is 1.0). When the configuration information 920 is used, the load when converted into the traffic amount 923 processed by the virtual server 403 is obtained. The calculated configuration information 920 is information set in advance from the client 101 or the like. The calculated configuration information 920 includes one entry including a virtual server 921 that stores the name of the virtual server 403, a placement server number 922 that stores an identifier of a physical server that provides the virtual server, and a processing traffic amount 923. The

  In FIG. 22, the CPU load of Node 0 is taken as an example. First, the configuration calculation unit 505 obtains “12%” by referring to the CPU load of Node 0 from the basic load per unit traffic in FIG. 20C. Next, the configuration calculation unit 505 refers to the information 920 regarding the configuration to be calculated in FIG. 22 and acquires traffic amount = 10000 to be processed by Node0. Since the node 0 has a CPU load of 12% per 1000 traffic, the configuration calculation unit 505 calculates that the CPU load is 120% when processing 10,000 traffic.

  Next, the configuration calculation unit 505 acquires “1” from the placement server number indicating the identifier of the physical server 202 that executes Node 0 from the information 920 regarding the configuration to be calculated in FIG. The configuration calculation unit 505 refers to the CPU performance 10013 of the physical server = “A” corresponding to the placement server number = “1” and acquires the CPU performance = “2.0”.

  In the case of the CPU performance 1.0, the configuration calculation unit 505 sets the node 0 to the physical server A having a CPU performance of 2.0 for the assumed load of the Node 0 in which the computational CPU load for processing 10,000 traffic is 120%. In the case of arrangement, the CPU load is calculated to be 60%. Then, the calculation result is stored in the assumed virtual server load in the middle of FIG. The configuration calculation unit 505 can calculate the assumed load of the physical server 202 from the average value of the assumed loads of the virtual server 403 arranged in the physical server 202. Similar to the CPU load, the configuration calculation unit 505 calculates a memory load, an I / O load, and a traffic load, and generates a virtual server assumed load 930 and a physical server assumed load 940 in FIG.

  The virtual server assumed load 930 is assigned to the virtual server name 931 that stores the name of the virtual server 403, the CPU load 932 that stores the expected value of the usage rate of the processor 220 assigned to the virtual server, and the virtual server. A memory load 933 for storing an estimated value of the usage rate of the memory 314, an I / O load 934 for storing an estimated value of a bandwidth usage rate of a storage device (not shown) accessed by the virtual server, and the virtual server One entry is composed of a traffic load 935 that stores an assumed value of the network load to be accessed.

  The assumed physical server load 940 includes a physical server name 941 for storing the name of the physical server 202, a CPU load 942 for storing an estimated value of the usage rate of the processor 220 of the physical server, and the memory 314 of the physical server. A memory load 943 for storing an assumed value of usage rate, an I / O load 944 for storing an assumed value of bandwidth usage rate of a storage device (not shown) accessed by the physical server, and a network I accessed by the physical server One entry is configured from the traffic load 945 that stores the expected value of the usage rate of / F401.

  When the virtual server assumed load 930 and the physical server assumed load 940 exceed preset default values (for example, 90%), the configuration calculation unit 505 determines the load of the corresponding virtual server and physical server. When the determination result 950 is set to Congestion, it is determined to be free if it is significantly lower than a preset specified value (for example, 20%), otherwise it is determined to be Normal, and the conditions_virtual_machine and free_virtual_machine Calculate the number. In FIG. 22, the load determination result 950 stores the determination result for each of the virtual server 403 and the physical server 202.

  Furthermore, all_virtual_machine in the above equation (1) represents the total number of virtual servers, and man_hour represents the man-hour for change. The change man-hour can be calculated by comparing genes.

  FIG. 23 shows a procedure by which the configuration calculation unit 505 calculates the change man-hour. In FIG. 23, the gene “10100233” when Node1 is connected to Virtual SW0 and the gene “100102233” when Node1 is connected to Virtual SW1 are compared, and the gene that is information indicating Node1 is compared. Since the second and third parts are different, the change man-hour is set to 1.

  In the above equation (1), as a weighting variable, A is a negative coefficient related to the load on the virtual server, B is a negative coefficient for the man-hour, a is a negative coefficient for a virtual server with a large load, and b is a virtual server with a small load. A negative coefficient, a position is a placement problem coefficient, a logical is a logical configuration coefficient, and “101002233” is set so that these coefficients are all between 0 and 1.

  Here, the weighting coefficients A, B, a, and b are determined based on the collected request information 803 and the request / constraint reflection table 1002. For example, when there is a request to reduce the man-hour as much as possible, B, which is a negative coefficient for the configuration, is set to be large according to the description in the request / constraint reflection table 1002. As the number of man-hours for change increases, the output value of the function of equation (1) becomes lower and is not calculated as a solution. This makes it possible to calculate a configuration that satisfies the user's request.

  In addition, when an individual (gene) that cannot satisfy the request information 803 is calculated, the fitness of the corresponding gene does not use the reward function (1), for example, by using a measure such as −1 of the fitness, Individual calculation that satisfies the requirements is easier to perform.

  Here, the threshold value of fitness is calculated from the value of urgency a which is a part of the input as shown in FIG. For example, when calculating the fitness threshold as shown in FIG. 18, the urgency a is 0 ≦ a ≦ 1, so that the higher the threshold, that is, the shorter the configuration calculation time, If the value decreases and the configuration calculated by the configuration calculation unit 505 satisfies the request to some extent, it is output as a solution. For this reason, the configuration calculation unit 505 can shorten the calculation time, and as a result, can quickly obtain the configuration calculation result.

  On the other hand, if the degree of urgency a is low, the threshold value of the fitness level increases, so the demand for the configuration calculated by the configuration calculation unit 505 becomes severe, and the time required to calculate the configuration becomes long. A configuration that satisfies the high demand is calculated as a solution. Although the conversion formula as shown in FIG. 18 does not have to be used for the calculation of the fitness, it is necessary to shorten the configuration calculation time if the urgency is high.

<Cluster adjustment>
Specifically, the cluster adjustment performed in steps S18 and S23 of FIG. 12 is executed by the configuration calculation unit 505 in a flowchart as shown in FIG.

  In FIG. 24, the configuration calculation unit 505 first corrects a configuration that does not satisfy the constraint conditions in step S51 using the lethal filter 1003 as in steps S13 and S15 of FIG. 12. Then, similarly to step S16 in FIG. 12, it is determined whether there is a configuration in which the fitness for each cluster exceeds the threshold (S52). If there is a configuration whose fitness level exceeds the threshold value, the cluster adjustment is terminated. On the other hand, if there is no configuration whose fitness exceeds the threshold value, the process proceeds to step S53.

  In step S53, the configuration calculation unit 505 obtains an average value of fitness for each cluster, and identifies a cluster having the lowest average value of fitness among clusters whose average value does not exceed the threshold value of fitness. If there is a cluster whose average fitness value is less than or equal to the threshold, the process proceeds to step S555, and if not, the process proceeds to step S54.

  In step S55, the cluster with the lowest average fitness value is the cluster with the highest possibility that the number of servers for each server type will not satisfy the user's request. Therefore, the configuration calculation unit 505 adds all the genes in the cluster. On the other hand, the ratio of the total number of virtual servers crowded over a certain number is obtained for Node, Virtual SW, and Virtual Router (step S5 in FIG. 24).

  This can be calculated by performing the same procedure (FIG. 22) as when congestion_virtual_machine was obtained during the evaluation of the configuration shown in FIG. 22.

  Here, if the percentage of genes whose total number of crowded virtual servers exceeds a certain value is large, the configuration calculation unit 505 determines that the number of physical servers 202 is insufficient.

 When the server type that has been insufficient is increased by one, a new cluster is created (S56). The cluster with the lowest fitness value gives the new cluster half its number of genes and halves its number of genes. Here, the configuration calculation unit 505 selects genes to be halved in descending order of fitness. If the number of genes halved is less than the specified number, all genes are given to the new cluster, and the original cluster disappears.

  The lower limit of the number of genes in the cluster may be set freely, but it is necessary to have at least two genes for crossover. In addition, if there is a new cluster created by the cluster with the lowest fitness average that has the same number for each server type, add half or all of the number of genes to the number of genes in the cluster. . A newly generated cluster or a cluster to which the number of genes is added generates a new gene for the added gene. In addition, a cluster for which no operation has been performed is performed as usual (S57).

  If all the clusters satisfy the prescribed fitness at the time of determining the average fitness in step S53, the cluster operation is not performed, and the process proceeds to step S54 to perform the normal selection operation.

  Then, after performing crossover in step S58, mutation is executed in step S59, the process returns to step S51, and the process is repeated until the fitness exceeds the threshold value in step S52. This operation makes it possible to search for the best number of servers for each server type that satisfies the user's request. A specific state of the cluster operation is shown in FIGS. 25 and 26 described later.

  FIG. 25 shows the cluster operation when a new cluster is created. Cluster A has 100 genes composed of 4 Nodes, 2 SWs, and 1 Router. Here, when the average fitness of cluster A is low and there are many genes indicating that the node is congested, that is, when the number of nodes seems to be insufficient, five nodes and two switches , A cluster B having a gene composed of a single router is created, and 50 genes that are half the number of genes in cluster A are newly created and held. Then, 50 of the genes in cluster A, 50, are deleted from those with low fitness.

  FIG. 26 shows a cluster operation when the number of genes is increased to an existing cluster. Cluster A has 100 genes composed of 4 Nodes, 2 SWs, and 1 Router. Cluster B has 50 genes composed of 5 Nodes, 2 SWs, and 1 Router. Here, when the average fitness of cluster A is low and there are many genes indicating that the node is congested, that is, when the number of nodes seems to be insufficient, five nodes and two switches , 50 genes that are half the number of genes in cluster A are added to cluster B, which has a gene that is composed of one Router. Cluster B, to which the number of genes has been added and the number of genes has been increased to 100, newly creates 50 genes and holds them. Then, 50 of the genes in cluster A, 50, are deleted from those with low fitness.

<Adaptability of physical configuration>
The calculation of the fitness of the physical configuration performed in step S21 in FIG. 12 is performed by the following equation, for example.

  When a logical configuration having a certain fitness level or more is calculated for all physical configurations, the fitness level of each physical configuration is calculated (step S21 in FIG. 12). The fitness calculation method is, for example,

Can be represented by

  Here, good_chemical_virtual_machine, bad_chemistry_virtual_machine, congestion_virtual_machine, and free_virtual_machine, all_virtual In addition, congestion_machine represents the number of physical servers having a high load, and free_machine represents the number of physical servers having a low load, and these variables can be obtained in the same manner as described above (see FIG. 22). Further, all_physical_machine in the above equation (2) represents the total number of physical servers, machine_num_i represents the number (i-th) for each physical server type, and machine_cost_i represents the cost (i-th) for each physical server type.

  FIG. 27 shows a procedure for obtaining the cost for each type of physical server. The cost for each type of the physical server 202 is obtained by comparing the information in the gene with the server information 1001 in the system definition information 804.

  In the example shown in FIG. 27, when referring to the gene expression, Machine type 0 and Machine 1 are expressed by the gene “10”, so that the server type A and Machine 2 are expressed by the gene “02”, so that the server type B As can be seen, the machine cost of Machine 0 and Machine 1 is 1.5, and the machine cost of Machine 2 is 1.0.

  Further, as a weighting coefficient, A is a negative coefficient related to the load on the virtual server, B is a negative coefficient relative to the man-hour, C is a negative coefficient related to the physical server load, D is a negative coefficient related to the physical server cost, and a is a virtual with a large load A negative coefficient for the server, b is a negative coefficient for a virtual server with a low load, c is a negative coefficient for a server with a large load, d is a negative coefficient for a server with a low load, further, position is a placement problem coefficient, logical is a logical configuration coefficient, “physical” is a physical configuration coefficient, and these coefficients are all set to values between 0 and 1. Here, as with the above-described function, the weighting coefficient is determined based on the collected request information 803 and the request / constraint reflection table 1002. For example, when there is a request to reduce the cost of the physical server as much as possible, the negative coefficient D relating to the cost of the physical server 202 is set to be large by the description of the request / constraint reflection table 1002. With this setting, the higher the cost of the physical server, the lower the output value of the above formula (1) or (2), and it is not calculated as a solution. This makes it possible to calculate a configuration that satisfies the user's request.

  Further, as described above, when there is an individual that satisfies a condition that is not permitted as a solution, the individual value that satisfies the request can be easily calculated by setting the reward value to be minus regardless of the reward function.

  FIG. 13 shows a modification of the present invention. FIG. 13 is a flowchart of the configuration calculation process performed by the calculation server 102 when the physical configuration is not considered. This flowchart is different from that shown in FIG. 12 and includes only the calculation part of the logical configuration and server arrangement, and the processing of the physical configuration in FIG. 12 is deleted. Therefore, a logical configuration that satisfies the determination in step S16 is output. This process corresponds to the flowchart B of the request / constraint reflection table 1002 in FIG. 18, and the process in FIG. 12 corresponds to the flowchart A of the request / constraint reflection table 1002.

  FIG. 28 shows a sequence diagram at the time of configuration change including change of the physical configuration. This sequence diagram is continued from S8 of the sequence diagram of FIG.

  First, the configuration presentation unit 506 receives the configuration (S61) calculated by the configuration calculation unit 505 in the calculation server 102 (S62). The configuration presentation unit 506 transmits the configuration change content to the client 101 via the network I / F 501 and the network 104. Then, the configuration change content transmitted by the network I / F 601 in the client 101 is received, and the configuration acquisition unit 602 stores the received configuration change content in the memory 607 as configuration information 901, and the configuration change content in the configuration display unit 603. (S63). The configuration display unit 603 that has received the storage completion notification of the configuration change content outputs the configuration information 901 in the memory 607 to the output device 612 via the I / O 606 and presents the change of the physical configuration to the administrator (user). (S64). The administrator changes the configuration of the physical server 202 or the like according to the presentation of the physical configuration.

  FIG. 29 is a sequence diagram at the time of configuration change that does not include a change in physical configuration. First, the configuration change instruction unit 507 receives the configuration (S71) calculated by the configuration calculation unit 505 in the calculation server 102 (S72). The configuration change instruction unit 507 transmits a configuration change request together with the configuration calculation result to the controller 201 in the data center 103 via the network I / F 501 and the network 104. When the configuration change request is received by the network I / F 303, the configuration change instruction unit 302 issues a configuration change command according to the request to the hypervisor 402 of the physical server 202 that requires the configuration change via the network I / F 301. (S73). After a sufficient time for the change to be reflected, has the configuration change instruction unit 302 been able to change the configuration to the hypervisor 402 of the physical server 202 that has changed the configuration via the network I / F 301? Confirmation is performed (S75, S76). When the configuration change has been performed as requested (S74), the configuration change instruction unit 302 notifies the calculation server 102 of the completion of the configuration change via the network 104 via the network I / F 301 and the router 203 (S77). When the notification of completion of the configuration change is received by the network I / F 501 in the calculation server 102, the configuration presentation unit 506 transmits the configuration change content to the client 101 via the network I / F 501 and the network 104 (S78). The configuration change content transmitted from the network I / F 601 in the client 101 is received (S 79), and the configuration acquisition unit 602 stores the received configuration change content in the memory 607 as configuration information 901, and the configuration display unit 603. Notify completion of storage of configuration changes. The configuration display unit 603 that has received the storage completion notification of the configuration change content presents the configuration information 901 in the memory 607 to the user via the I / O 606 (S80).

  If the physical configuration is not changed by the above processing, the configuration of the virtual server 403 is changed by the hypervisor 402.

  As described above, according to the present invention, by considering the connection relationship between the physical servers 202, the connection relationship between the virtual servers 403 in the virtual machine system, and the arrangement relationship between the virtual server 403 and the physical server 202, It becomes possible to quickly calculate a configuration with a high degree of freedom that could not be realized. Further, since the user of the client 101 can calculate the configuration of the entire virtual machine system simply by inputting needs (request information), it is not necessary for the user to consider the configuration with a great deal of effort, and management is performed as a result. Costs can be reduced.

  In the above embodiment, an example of obtaining the configuration pattern of the components of the virtual computer system that configures the data center 103 has been shown. However, only the arrangement pattern of the virtual server or only the connection pattern of the physical server 202 and the switch is used. You may make it calculate.

101 Client 102 Computing Server 103 Data Center 104 Network 201 Controller 202 Physical Server 203 Router 204 Switch 301 Network I / F
302 Configuration change instruction unit 303 Configuration / load information acquisition unit 304 Memory 401 Network I / F
402 Hypervisor 403 Virtual server 501 Network I / F
502 Request acquisition unit 503 Configuration / load information acquisition unit 504 Memory 505 Configuration calculation unit 506 Configuration presentation unit 507 Configuration change instruction unit 601 Network I / F
602 Configuration acquisition unit 603 Configuration display unit 604 Request acquisition unit 605 Request transmission unit 606 I / O
607 Memory 701 Configuration information 702 Load information 801 Configuration information 802 Load information 803 Request information 804 System definition information 901 Configuration information 902 Request information 1001 Server information 1002 Request / constraint reflection table 1003 Lethal filter

Claims (20)

A management computer including a processor and a memory calculates a configuration pattern of elements constituting a computer system,
A first step in which the management computer obtains configuration information including information on elements constituting the computer system;
A second step in which the management computer obtains load information including a load of an element constituting the computer system;
A third step in which the management computer obtains a constraint condition for a configuration pattern of the computer system;
A fourth step in which the management computer generates a plurality of configuration patterns based on configuration information of the computer system;
A fifth step in which the management computer corrects a configuration pattern that does not satisfy the constraint condition among the generated plurality of configuration patterns;
A sixth step in which the management computer calculates a predetermined fitness for each of the configuration pattern that satisfies the constraint condition and the modified configuration pattern by referring to the configuration information and load information;
Said management computer, if the fitness is present arrangement pattern exceeds a predetermined threshold, a seventh step of outputting the configuration pattern,
Said management computer, if the fitness is not present configuration pattern exceeds a predetermined threshold value, the fitness and the step of deleting the configuration pattern below a predetermined threshold value, one configuration pattern and another certain configuration pattern An eighth step of applying a step of replacing a part and a step of changing a part of a certain configuration pattern;
Returning to the fifth step after the eighth step ,
The first step includes
Obtaining configuration information including physical components of the computer system and information of logical components;
The second step includes
Obtaining load information including a physical component load and a logical component load of the computer system;
The fourth step includes
Generating a configuration pattern of physical components of the computer system and a configuration pattern of logical components;
The fifth step includes
The physical configuration pattern that does not satisfy the constraint condition among the plurality of generated physical configuration patterns is corrected, and the logic that does not satisfy the constraint condition among the plurality of generated logical configuration patterns The typical composition pattern
The sixth step includes
Referring to the physical configuration information and the physical load information, a predetermined fitness is calculated for each of the physical configuration pattern that satisfies the constraint condition and the modified physical configuration pattern, and the logic A predetermined fitness for each of the logical configuration pattern that satisfies the constraint condition and the modified logical configuration pattern, with reference to the general configuration information and the logical load information,
The seventh step includes
When there is a physical configuration pattern in which the fitness exceeds a predetermined threshold, and there is a logical configuration pattern in which the fitness exceeds a predetermined threshold, the physical configuration pattern and logical A calculation method for a configuration pattern of a computer system, characterized in that a simple configuration pattern is output . A method for calculating a configuration pattern of a computer system according to claim 1,
The fourth step includes
A method for calculating a configuration pattern of a computer system, comprising generating the configuration pattern converted into a numerical sequence indicating a combination of elements constituting the computer system. A method for calculating a configuration pattern of a computer system according to claim 2,
The fourth step includes
A method for calculating a configuration pattern of a computer system, wherein sequences having different lengths depending on the number of components of the configuration pattern and having the same length are handled as one cluster. A method for calculating a configuration pattern of a computer system according to claim 3,
The seventh step includes
If the fitness in the cluster are present arrangement pattern exceeds a predetermined threshold value, and outputs the configuration pattern,
The eighth step, when the fitness of the said cluster no configuration pattern exceeds a predetermined threshold, the steps of the fitness deletes the configuration pattern below a predetermined threshold value, one configuration pattern and another A method for calculating a configuration pattern of a computer system, wherein a step of replacing a part of a certain configuration pattern and a step of changing a part of the configuration pattern are applied. A method for calculating a configuration pattern of a computer system according to claim 2,
The fourth step includes
In the numerical sequence, a different numerical value is set for each element of a different type. A method for calculating a configuration pattern of a computer system according to claim 1,
A number sequence indicating a combination of elements constituting the computer system is a gene,
The fourth step includes
Depending on the number of components of the configuration pattern, the length of the sequence is different, and the sequences having the same length of the sequence are treated as one cluster,
Generating an initial pattern of the constituent pattern of the physical constituent elements in a gene pattern for each cluster; and
If the generated configuration pattern of the physical component is contrary to the constraint condition, correcting the gene of the physical component configuration pattern so as to satisfy the constraint condition;
Generating an initial pattern of a configuration pattern of the logical component with a gene pattern for each cluster;
When the gene of the generated configuration pattern of the logical component is contrary to the constraint condition, correcting the gene so as to satisfy the constraint condition;
A method for calculating a configuration pattern of a computer system, comprising: A configuration pattern calculation apparatus that calculates a configuration pattern of elements that comprise a computer system and includes a processor and a memory,
A configuration information acquisition unit for acquiring configuration information including information of elements constituting the computer system;
A load information acquisition unit for acquiring load information including a load of elements constituting the computer system;
A request acquisition unit for acquiring a constraint condition for a configuration pattern of the computer system;
A configuration calculation unit that generates a plurality of configuration patterns based on the configuration information of the computer system,
The configuration calculation unit
Of the plurality of generated configuration patterns, a correction unit that corrects a configuration pattern that does not satisfy the constraint condition;
An fitness calculation unit that calculates a predetermined fitness for each of the configuration pattern that satisfies the constraint condition and the modified configuration pattern with reference to the configuration information and the load information;
If the fitness is present arrangement pattern exceeds a predetermined threshold value, the configuration presentation unit to output the configuration pattern,
Step the fitness in the absence of structure pattern exceeds a predetermined threshold value, interchanging the steps of the fitness deletes the configuration pattern below a predetermined threshold, the part of the configuration pattern and another certain configuration patterns that And a configuration changing unit that applies a step of changing a part of a certain configuration pattern,
The configuration pattern calculated by the configuration calculation unit in the configuration change unit and output by the configuration presentation unit is input to the correction unit ,
The configuration information acquisition unit
Obtaining configuration information including physical components of the computer system and information of logical components;
The load information acquisition unit
Obtaining load information including physical component load and logical component load of the computer system;
The configuration calculation unit
Generating a configuration pattern of physical components of the computer system and a configuration pattern of logical components;
The correction unit is
The physical configuration pattern that does not satisfy the constraint condition among the plurality of generated physical configuration patterns is corrected, and the logic that does not satisfy the constraint condition among the plurality of generated logical configuration patterns The typical composition pattern
The fitness calculator is
Referring to the physical configuration information and the physical load information, a predetermined fitness is calculated for each of the physical configuration pattern that satisfies the constraint condition and the modified physical configuration pattern, and the logic A predetermined fitness for each of the logical configuration pattern that satisfies the constraint condition and the modified logical configuration pattern, with reference to the general configuration information and the logical load information,
The configuration presentation unit
When there is a physical configuration pattern in which the fitness exceeds a predetermined threshold, and there is a logical configuration pattern in which the fitness exceeds a predetermined threshold, the physical configuration pattern and logical A configuration pattern calculation apparatus that outputs a configuration pattern. The apparatus for calculating a configuration pattern according to claim 7,
The configuration calculation unit
An apparatus for calculating a configuration pattern of a computer system, which generates the configuration pattern converted into a numerical sequence indicating a combination of elements constituting the computer system. The apparatus for calculating a configuration pattern according to claim 8,
The configuration calculation unit
An apparatus for calculating a configuration pattern, characterized in that sequences having different lengths depending on the number of components of the configuration pattern and having the same length are handled as one cluster. The apparatus for calculating a configuration pattern according to claim 9,
The configuration presentation unit
If the fitness in the cluster are present arrangement pattern exceeds a predetermined threshold value, and outputs the configuration pattern,
The configuration change unit, when the fitness of the said cluster no configuration pattern exceeds a predetermined threshold, the steps of the fitness deletes the configuration pattern below a predetermined threshold value, one configuration pattern and another An apparatus for calculating a configuration pattern, characterized by applying a step of replacing a part of a certain configuration pattern and a step of changing a part of the configuration pattern. The apparatus for calculating a configuration pattern according to claim 8,
The configuration calculation unit
In the numerical sequence, a different numerical value is set for each element having a different type. The apparatus for calculating a configuration pattern according to claim 7,
The configuration calculation unit
A number sequence indicating a combination of elements constituting the computer system is a gene,
Depending on the number of components of the configuration pattern, the length of the sequence is different, and the sequences having the same length of the sequence are treated as one cluster,
When an initial pattern of a physical component configuration pattern is generated as a gene pattern for each cluster, and the generated physical component configuration pattern violates the constraint condition, The gene of the component pattern of the general component is modified so that the constraint condition is satisfied, the initial pattern of the component pattern of the logical component is generated as the gene pattern for each cluster, and the generated logical pattern An apparatus for calculating a configuration pattern , wherein when a gene of a configuration pattern of a component violates the constraint condition, the gene is corrected so as to satisfy the constraint condition .   A management computer including a processor and a memory calculates a configuration pattern of elements constituting a computer system,
  A first step in which the management computer obtains configuration information including physical components of the computer system and logical component information;
  A second step in which the management computer obtains load information including a physical component load of the computer system and a logical component load;
  A third step in which the management computer obtains a constraint condition for a configuration pattern of the computer system;
  A fourth step in which the management computer generates a configuration pattern of a physical component of the computer system and a configuration pattern of a logical component based on the configuration information of the computer system;
  The management computer corrects a physical configuration pattern that does not satisfy the constraint condition among the plurality of generated physical configuration patterns, and the constraint condition among the generated plurality of logical configuration patterns. A fifth step of correcting a logical configuration pattern that does not satisfy
  With reference to the logical configuration information and the logical load information, a predetermined fitness is calculated for each of the logical configuration pattern satisfying the constraint condition and the modified logical configuration pattern, and the adaptation A sixth step of determining whether there is a logical configuration pattern whose degree exceeds a predetermined threshold;
  If there is a logical configuration pattern whose fitness exceeds the predetermined threshold, the physical condition satisfying the constraint condition is referred to by referring to the physical configuration information and physical load information. A predetermined fitness is calculated for each of the correct configuration pattern and the modified physical configuration pattern, and a seventh determination is made as to whether or not there is a physical configuration pattern with the fitness exceeding a predetermined threshold. Steps,
  An eighth step in which the management computer outputs the physical configuration pattern and the logical configuration pattern when there is a physical configuration pattern whose fitness exceeds a predetermined threshold; and
  The management computer, when there is no physical configuration pattern in which the fitness exceeds a predetermined threshold, deleting a physical configuration pattern in which the fitness is less than a predetermined threshold; and a configuration pattern A ninth step of applying a step of replacing a part of another certain physical configuration pattern and a step of changing a part of the certain physical configuration pattern;
And a return to the fifth step after the ninth step. A configuration pattern calculation apparatus that calculates a configuration pattern of elements that comprise a computer system and includes a processor and a memory,
  A configuration information acquisition unit for acquiring configuration information including physical components constituting the computer system and information on logical components;
  A load information acquisition unit for acquiring load information including a load of physical components constituting the computer system and a load of logical components;
  A request acquisition unit for acquiring a constraint condition for a configuration pattern of the computer system;
  A configuration calculation unit that generates a configuration pattern of physical components and a configuration pattern of logical components based on the configuration information of the computer system, and
  The configuration calculation unit
  The physical configuration pattern that does not satisfy the constraint condition among the plurality of generated physical configuration patterns is corrected, and the logic that does not satisfy the constraint condition among the plurality of generated logical configuration patterns A correction unit for correcting a typical configuration pattern;
  With reference to the logical configuration information and the logical load information, a predetermined fitness is calculated for each of the logical configuration pattern satisfying the constraint condition and the modified logical configuration pattern, and the adaptation When there is a logical configuration pattern whose degree exceeds a predetermined threshold, the physical configuration pattern satisfying the constraint condition and the modified configuration pattern are referred to by referring to the physical configuration information and physical load information. Fitness calculation unit for calculating a predetermined fitness for each physical configuration pattern,
  When there is a physical configuration pattern whose fitness exceeds a predetermined threshold, a configuration presentation unit that outputs the physical configuration pattern and the logical configuration pattern;
  When there is no physical configuration pattern whose fitness exceeds a predetermined threshold, a step of deleting a physical configuration pattern whose fitness is lower than a predetermined threshold; A configuration changing unit that applies a step of replacing a part of a certain physical configuration pattern and a step of changing a part of a certain physical configuration pattern;
  A configuration pattern calculation apparatus, wherein the configuration change unit inputs a logical configuration pattern calculated by the configuration calculation unit and a physical configuration pattern output by a configuration presentation unit to the correction unit.   A method for calculating a configuration pattern of a computer system according to claim 1,
  The method of calculating a configuration pattern of a computer system, wherein the fitness is an urgency level that is preset and determines the calculation accuracy of the configuration pattern.   A method for calculating a configuration pattern of a computer system according to claim 15,
  The threshold value is calculated based on the urgency level, and when the urgency level is high, the threshold value is decreased to reduce the calculation time until the configuration pattern is output, and when the urgency level is low. A method for calculating a configuration pattern of a computer system, characterized in that the calculation time until the threshold value is set high and the configuration pattern is output is increased.   A method for calculating a configuration pattern of a computer system according to claim 1,
  The physical load information obtains a load of a physical server that is the physical configuration information,
  The logical load information obtains a load of a virtual server which is the logical configuration information, and a computer system configuration pattern calculation method.   An apparatus for calculating a configuration pattern of a computer system according to claim 7,
  An apparatus for calculating a configuration pattern of a computer system, wherein the fitness is an urgency level that is preset and determines the calculation accuracy of the configuration pattern.   A calculation device for a configuration pattern of a computer system according to claim 18,
  The threshold value is calculated based on the urgency level, and when the urgency level is high, the threshold value is decreased to reduce the calculation time until the configuration pattern is output, and when the urgency level is low. Is a configuration pattern calculation apparatus for a computer system, characterized in that the calculation time until the threshold value is set high and the configuration pattern is output is increased.   An apparatus for calculating a configuration pattern of a computer system according to claim 7,
  The physical load information obtains a load of a physical server that is the physical configuration information,
  The logical load information acquires a load of a virtual server, which is the logical configuration information, and a computer system configuration pattern calculation apparatus.