JP2007200128A - Computer system, management server, method for reducing computer setup time, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce a computer setup time while suppressing the number of necessary computers. <P>SOLUTION: A state determination part 112 determines an optimum state mainly about a computer resource on the basis of state transition rule information acquired from a state transition rule DB 121, a state determination index information acquired from a state determination index DB 122, a state transition prediction time acquired from a time prediction part 113, and a present state of a resource acquired from a state acquisition part 114. The optimum state is a state having a minimum expectation of a resource setup time, or an expectation close to a minimum. The optimum state determined by the state determination part 112 is delivered to a transition instruction part 115. The transition instruction part 115 changes the state of the resource, or performs an instruction of the change to an agent 201 on each computer so as to bring the computer into the optimum state determined by the state determination part 112. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a computer system, a method for reducing the computer setting time, and a program for reducing the computer setting time for reducing the computer setting time. In particular, the computer setting time is reduced in a situation where a plurality of procedures are required for computer setting. The present invention relates to a computer system capable of reducing, a method of reducing computer setting time, and a program for reducing computer setting time.

  Companies that provide services using computers need to have software and computers that can withstand the maximum demands in order to cope with the increase in service demands. However, there is a problem that a fixed cost is required to maintain software and computers that can withstand the maximum requirements. In order to improve this problem, it is preferable that necessary resources such as software and computers can be temporarily added when there is a sudden increase in service demand. Therefore, in recent years, utility computing that contracts and uses services, software, computers, and the like using computers for the purpose of reducing fixed costs and dealing with a sudden increase in service demand has attracted attention.

  Since there is a service whose service request amount varies greatly with time, utility computing is required to cope with a sudden increase in service request. For example, in a web page providing service for a specific event such as a baseball game, the service request amount increases abruptly when the event is started compared to before the event is started. If it cannot cope with the rapid increase in the amount of service requests, the satisfaction level of general users who receive web pages may be reduced, leading to a decrease in general users.

  In order to cope with a sudden increase in service request amount, it is necessary to increase the amount of resources such as the number of computers when the service request amount increases. For example, if the number of computers required in proportion to the service request amount is increased, if the service request amount increases, the number of computers needs to be increased in proportion thereto.

  The problem here is that it takes time to set up the computer, so it cannot respond quickly to an increase in the amount of service requests. In the setting of the computer, for example, operations such as OS (Operating System) installation, OS setting, middleware installation, middleware setting, application installation, application setting, application activation, and data copying are required. It may take several hours to perform all these operations. Therefore, if these operations are performed after the service request amount has increased, many services cannot cope with the increase in the service request amount.

  Two methods have been used to address this problem. The first method is a method of setting a computer and maintaining the computer in a state (hot standby) that can respond to a service in a short time. The second method is a method of predicting an increase in the required amount of service in advance and setting the computer in advance in accordance with the predicted required amount. The first method has the following problems.

  The hot standby means a computer in the same state as the computer that provides the service or a computer in that state. A feature of a computer in a hot standby state is that setting for a service can be completed in a very short time.

  The problem with hot standby is that many standby computers are required when there are many targeted services. In a system using hot standby, as many computers as necessary are prepared for each target service. Therefore, it is necessary to prepare a number of standby computers proportional to the number of target services. When the number of targeted services increases, the number of necessary standby computers increases, and the cost for the computers increases accordingly. Therefore, the fixed cost reduction that is the purpose of utility computing cannot be achieved. It is also conceivable that all standby computers are set to hot standby for a specific service as much as possible. However, in such a case, when a use request for a service for which a hot standby is not prepared is received, there is a problem that it takes a long time to reset the hot standby for another service.

  Non-Patent Document 1 describes a technique for shortening a computer setting time (hereinafter also referred to as a computer resource setting time) using a method called a resource cache. The resource cache is a method of making a software ready for use without erasing software on a computer that no longer uses the service. A computer that is ready to use software is called a cached computer. When service requests increase again due to the resource cache, the resource cached computer is set to an immediately available state.

  A problem with the resource cache is that there is a large penalty when a cache miss occurs. A cache miss means that a cached resource is needed for a service that cannot be added immediately. If it takes time to return the cached resource state to the initial state, the resource setting time becomes very long. In this case, there arises a problem that it is difficult to quickly cope with an increase in service demand. This problem is similar to the hot standby problem.

  In Non-Patent Document 2, data necessary for a service is not transferred at the time of initialization by preparing an OS necessary for the service to use a computer and using NFS (Network File System). Thus, a method for shortening the preparation time of software and computers is described. This method does not make it possible to add a computer immediately.

  Problems with the method described in Non-Patent Document 2 include that the setting time cannot be said to be sufficiently short depending on the service request, and that this method cannot be used depending on the service. In the method described in Non-Patent Document 2, each set time is shortened, but whether or not the time is sufficient depends on a service request. In addition, since NFS is used, the overhead at the time of execution is large, and it may be unsuitable for applications using a large amount of data.

  As a technique related to the second method of predicting an increase in service request amount in advance and setting a computer in advance in accordance with the predicted request amount, Patent Document 1 discloses a change in service request amount. A system is described that transitions the state between a hot standby state and a state in which no application is entered, depending on the prediction. In the system described in Patent Document 1, the standby computer resource has at least a dead standby state in which no application is installed, and the computer resource in the dead standby state is shared by a plurality of services and a plurality of users. As a result, the utilization rate of idle computer resources and server integration are realized, and the cost required for maintaining computer resources is reduced. Also, idle computer resources are secured from surplus services, load prediction is performed using past operation history for each service, and idle computer resources are input according to the prediction result.

  Here, dead standby is a state in which at least an application is not installed, and can be a standby system for any service by installing an application or OS.

  However, since Patent Document 1 defines only hot standby and dead standby states, it can be considered that there are two problems.

  The first problem is that if all idle computer resources are placed in hot standby according to the load prediction when the number of idle computer resources is small, a problem similar to the hot standby problem occurs. That is, when many computer resources are required for load prediction, it is conceivable that the computer resources are insufficient.

  The second problem is that when many idle computer resources are maintained as a dead standby, it takes a long time to set a service available state from the dead standby, and the load prediction error may increase. The error of load prediction when predicting the load state after a long time may be larger than the error of load prediction when predicting the load state after a short time. When the error in the load prediction becomes large, problems such as being unable to respond to service requests and resources being wasted arise.

JP 2005-141605 A Karen Appleby and eight other authors, Oceano-SLA Based Management of a Computing Utility, IFP / IEE IF International Symposium on Integrated Network Management (IFIP / IEEE International Symposium on Integrated Network Management), USA, May 2001 Mineyoshi Masuda and 4 other authors, VPDC Virtual Private Data Center A Flexible and Rapid Workload-Management System (VPDC), IF・ IP / IEEE International Symposium on Integrated Network Management (IFIP / IEEE International Symposium on Integrated Network Management), USA, May 2003

  The first problem is that, when the conventional setting method for shortening the setting time by hot standby is applied, the number of necessary computer resources increases in a situation where there are many services to be processed. The reason is that the conventional computer setting time reduction method requires a hot standby for each service, so the number of required computers increases in proportion to the number of services to be set. is there.

  The second problem is that when all standby computers are set as hot standby for a specific service as much as possible, when a use request for a service that does not have hot standby is received, It takes a long time to reset hot standby. The reason is that with the conventional computer setting time reduction method, if a service use request that does not have a hot standby is received, the setting must be made from scratch, and the time is increased accordingly. is there.

  The third problem is that when many idle computer resources are maintained as a dead standby, it takes a long time to set the service to a usable state from the dead standby, and the load prediction error may increase. It is. This is because the load prediction error for predicting the load state after a long time may be larger than the load prediction error for predicting the load state after a short time.

  An object of the present invention is to reduce the computer setting time while suppressing the number of necessary computers. Another object of the present invention is to averagely reduce the computer setting time after receiving a computer resource utilization request in a situation where the number of standby computers is small. Still another object of the present invention is to reduce load prediction errors by keeping the computer setting time short.

  Still another object of the present invention is to provide a computer setting time reduction method that does not depend on a specific technique. Note that the computer setting time can be reduced as compared with the conventional method by using a method for reducing the computer setting time that does not depend on a specific technique in combination with a method that depends on the specific technique.

  The computer system according to the present invention uses the predicted value of the set time, which is the time required for setting the computer resource, and the computer resource setting procedure to determine the state of the computer resource where the expected value of the set time is reduced. It is characterized by comprising. Note that the state of the computer resource in which the expected value of the set time is small is, for example, a state in which the set time is set to be shorter than in the case of using the above-described conventional technology. This is the state of computer resources that minimizes the expected value of. In another aspect of the present invention, the computer system further includes state transition means for presetting the computer resources to the state determined by the state determination means. In the present invention, the setting time of the computer resource can be shortened on average by maintaining the state of the computer resource in a hot standby state, a dead standby state, and an intermediate state thereof.

  The first effect is that the setting time of computer resources can be shortened while reducing the number of necessary computers. The reason is that the number of idle computers is taken into consideration and the state is set so as to be allocated to a plurality of services in a short time. When determining the state, by using the predicted value of the setting time of the computer, an optimal state that shortens the setting time of the computer resource can be obtained.

  The second effect is that the setting time of computer resources can be shortened on average in a situation where the number of standby computers is small. The reason is that the average setting time can be shortened by maintaining the computer resources in such a state that a plurality of services can be immediately set to be usable. What state the computer resource should be in is determined by calculation, and the computer resource is set to that state. As a result, the average setting time can be shortened compared to a system using only hot standby.

  A third effect is that, when performing resource setting based on service load prediction, there is a possibility that an error in the load prediction can be reduced. The reason is that by maintaining the computer resources in a state that can be set in a shorter time than the dead standby, the time point when the load prediction is performed can be made closer.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

Embodiment 1 FIG.
In the first embodiment, the computer resource setting time is reduced by setting the computer resource in advance. Here, the setting of the computer resource includes the setting of the network related thereto, the setting of the application, the setting of the user (requester of the computer resource), and the like. In the computer system of this embodiment, it is assumed that the service to be used and the service setting procedure are known in advance. Moreover, the setting procedure of a computer resource and the state transition of a computer resource are used in a substantially similar meaning. More specifically, a computer resource setting procedure means a procedure for making a specific service available on a computer resource. The state transition of the computer resource means that the setting on the computer resource has been changed and the state of the computer resource has changed. Computer resources and resources have the same meaning. A computer resource or resource means a software component on a computer including a computer, a network, and a service component.

  FIG. 1 is a block diagram illustrating a computer system according to the first embodiment. As shown in FIG. 1, the computer system according to the first embodiment of this invention includes a management server 100, computers 102, 103, and 104 that are connected to the management server 100 via a network 101 and perform calculations for providing services. Including.

  FIG. 2 is a block diagram showing a configuration of a computer for providing a service. The computers 200 and 210 shown in FIG. 2 correspond to the computers 102, 103 and 104 shown in FIG. 1, but only two are shown in FIG. , The reference numerals in FIG. 1 are changed. In FIG. 2, a computer 200 shows a state where the OS is already installed, and a computer 210 shows a state where the OS has not been installed yet.

  The computer 200 includes an agent 201 that operates according to an instruction from the management server 100 and a service component 202 that provides some service. In response to an instruction from the management server 100, the agent 201 acquires the state of the computer 200, sets a service component, and acquires a set time. However, if the computer does not include an OS (Operating System) as in the computer 210, there is no agent. In this situation, OS installation and agent installation work are performed from the management server 100.

  The service component 202 has a role of providing a service on each computer. There may be multiple service components. The service component receives operations such as installation, setting, and activation from the agent.

  FIG. 3 is a block diagram illustrating a configuration of the management server 100. The management server 100 includes a data processing device 110 and a storage device 120. The data processing device 110 includes a start unit 111, a state determination unit 112, a time prediction unit 113, a state acquisition unit 114, and a transition instruction unit 115. The start unit 111, the state determination unit 112, the time prediction unit 113, the state acquisition unit 114, and the transition instruction unit 115 are implemented by a CPU of the data processing apparatus 110 that executes a program and a process according to the program. The state determination unit 112 is an example of a state determination unit that determines the state of a computer resource that reduces the expected value of the set time. The transition instruction unit 115 presets the computer resource to the state determined by the state determination unit. It is an example of the state transition means to do. The time prediction unit 113 is an example of a state transition time prediction unit that predicts a set time that is a time required for setting a computer resource and generates a predicted value of the state transition time.

Each component in the management server 100 operates as follows.
The starting unit 111 determines a change in resource state, periodic resource state transition, a service user request, and the like, and determines whether or not to start the resource state determination process according to the determination result. If it is determined to start the resource status determination process, an instruction to that effect is issued to the status determination unit 112. In the following description, a service user means a person who requests a computer resource from a computer system and receives the computer resource. A service means that a service user who receives the provision of computer resources is allowed to use a service received from an operator of the computer system, for example, a computer in which an application desired by the service user can be used. Means that.

  When receiving an instruction from the state determination unit 112, the state determination unit 112 acquires the state transition rule information acquired from the state transition rule DB (database) 121, the state determination index information acquired from the state determination index DB 122, and the time prediction unit 113. Based on the predicted state transition time and the current state of the resource acquired from the state acquisition unit 114, an optimal state is determined mainly for the computer resource. Here, the optimal state is a state in which the expected value of the resource setting time determined according to the state determination index is minimum or close to the minimum. The optimum state determined by the state determination unit 112 is passed to the transition instruction unit 115.

  The time prediction unit 113 receives a request from the state determination unit 112 and predicts the state transition time. As a method of predicting the state transition time, a method of predicting from a past state transition history, a method of predicting from a modeled state transition, a method of predicting an entire state transition time from a partial execution time, or a combination of these methods Can be used.

  The method of predicting from the past state transition history is a method of calculating the predicted time from the average value of the past state transition history or calculating the predicted time using another statistical method. The method of predicting from a modeled state transition is a method of calculating a prediction time by modeling a state transition as an expression using a computer attribute as a variable and applying an actual state to the variable. The method of predicting the entire state transition time from the partial execution time is a method of executing a part of the state transition, estimating the entire execution time from the execution time, and calculating the predicted time. The calculated state transition prediction time is passed to the state determination unit 112 that is a request source.

  The state acquisition unit 114 acquires the current state. The current state is a state where the resource is located at that time on the state transition rule. The timing at which the state acquisition unit 114 acquires the state may be acquired after receiving the state acquisition request, or may be acquired in advance before receiving the state acquisition request. A description of the state transition rules is detailed below.

  The transition instruction unit 115 changes the resource state or instructs the agent 201 on each computer to change the state of the resource in order to bring the computer into the optimum state determined by the state determination unit 112. The changes performed by the transition instruction unit 115 are OS installation, agent 201 installation, service component 202 setting instruction, and other resource state changes.

  Next, the configuration of the storage device 120 shown in FIG. 3 will be described. The storage device 120 includes a state transition rule DB 121, a state determination index DB 122, and a transition time history DB 123.

  The state transition rule DB 121 holds a rule for transition of the resource state. The state transition rule is composed of a set of states and a set of branches connecting the states. Each branch has a direction. A state can only transition to the state indicated by the branch. If it is possible to go back and forth between the two states, there will be two branches facing in opposite directions. The starting point of state transition is a state where nothing is set for each resource. For example, in a computer resource, a state in which neither an OS nor an agent is included is a starting point of state transition. The state transition rule DB 121 passes the state transition rule held to the state determination unit 112.

  In the present embodiment, the state transition rule is stored in advance in the state transition rule DB 121. The state transition rule is set in advance by a system administrator or the like, for example. A specific example of the state transition rule will be described in detail in the first embodiment. Further, a method for simplifying the input of the state transition rule is realized in the third embodiment.

  Next, the state determination index DB 122 will be described. The state determination index DB 122 holds an index (state determination index) for determining the resource state. The state determination index is, for example, the arrival probability of a resource setting request (a request that a service user requests to set a resource for a specific service), the priority of a service, or the priority for each service user. Degree. The state determination index is used by the state determination unit 112.

  The transition time history DB 123 holds a history of the time taken for each state transition. The history is used by the time prediction unit 113 to predict the state transition time.

  Next, the overall operation of the present embodiment will be described with reference to the flowcharts of FIGS. 4, 5, 6, and 7. FIG. 4 is a flowchart showing the flow of the state determination operation. FIG. 5 is a flowchart showing the flow of the transition time prediction operation by the time prediction unit 113. FIG. 6 is a flowchart showing the flow of the resource status acquisition operation by the status acquisition unit 114. FIG. 7 is a flowchart showing the flow of the state transition execution operation by the transition instruction unit 115.

  When the processing is started by the start unit 111, that is, when the start unit 111 issues a start request, the state determination unit 112 that has received the start request acquires a state transition rule from the state transition rule DB 121 (step A1). Next, the state determination part 112 acquires the estimated value of the time concerning each state transition from the time prediction part 113 according to a state transition rule (step A2). The start unit 111 issues a start request to the state determination unit 112 when a service use request is received from a service user terminal device via a communication network such as the Internet.

  The operation of the time prediction unit 113 will be described with reference to FIG. The time prediction unit 113 receives a time prediction request from the state determination unit 112 (step B1). Receiving the time prediction request, the time prediction unit 113 inputs the history of state transition time from the transition time history DB 123 (step B2). Then, the state transition time is predicted using the input state transition time (step B3). The prediction method includes a method of predicting from the past state transition history, a method of predicting from the modeled state transition, a method of predicting the entire state transition time from the partial execution time, and those methods. There are various methods besides the combined method. The time prediction unit 113 outputs the obtained state transition prediction time to the state determination unit 112 (step B4).

  The state determination unit 112 that has input the state transition rule and the state transition prediction time acquires the state determination index from the state determination index DB 122 (step A3). Furthermore, the state determination unit 112 acquires the resource state from the state acquisition unit 114 (step A4).

  The state acquisition process will be described with reference to FIG. First, the state acquisition unit 114 receives a resource state acquisition request from the state determination unit 112 (step C1). Next, the state acquisition unit 114 acquires the current resource state (step C2). The status acquisition unit 114 may query the agent on each computer for resource status information after receiving the resource status acquisition request, or acquire the resource status information in advance before receiving the resource status acquisition request. You may keep it. Next, the state acquisition unit 114 outputs the acquired resource state to the state determination unit 112 (step C3).

  Note that the process of step A1 should be executed before the process of step A2, but the process of step A3 and the process of step A4 may be executed before the process of step A1, or the process of step A4 The process of step A3 may be executed before the process.

  Next, the state determination unit 112 determines the optimal state of the resource using the resource information acquired in Step A4 (Step A5). The optimum state is, for example, a state in which the expected value of the resource setting time is minimized, or a state according to the priority when priority is given to each service or user. In addition, as a method for determining the state, a full search method for determining whether each combination of possible states is optimal may be used. If the total search method requires a large amount of calculation, a combination optimization method may be used. May be used.

  Thereafter, the state determination unit 112 compares the determined optimal state with the current state (step A6), and if the state transition is not necessary, the state determination unit 112 ends. If state transition is required, the transition instruction unit 115 is requested to perform state transition (step A7).

  The operation of the transition instruction unit 115 will be described with reference to FIG. First, the transition instruction unit 115 receives a state transition instruction from the state determination unit 112 (step D1). Next, the resource which performs a state transition is selected (step D2), the state transition is instructed to the agent, or the state transition is executed by itself (step D3). Thereafter, the state determination unit 112 is notified of whether the state transition has succeeded or failed (step D4).

  In the present embodiment, when a computer resource use request is made, a computer resource use requester, that is, a service user selects a resource. As the resource selection index, among the computer resources that meet the conditions, the resource setting time is sufficiently short, or the cost for use is low. Taking the case where the resource selection index is the setting time of a computer resource as an example, the computer resource use requester first displays a list of available resources before the use requester requests the resource use request. A request is made to the management server 100 using a terminal device or the like that has it. The resource selection unit (not shown in FIG. 3) of the management server 100 sets the computer resources required for setting the requested service from the current state of each resource until the requested service can be provided to the use requester. The time is predicted by the time prediction unit 113. Then, in order from the shortest setting time of the computer resource predicted by the time prediction unit 113, a resource list in which information that can identify the resource, for example, an ID attached to the resource is set, is created. The resource list is transmitted to a terminal device or the like possessed by the use requester. In the resource list, in addition to information that can identify the resource, the performance of the resource and the set time of the computer resource are also set. The use requester selects a resource by looking at the resource list, and makes a use request for the selected resource to the management server 100. In general, the use requester selects a resource with a short computer resource setting time. Therefore, the use requester can use the resource in a short time. When the number of free resources is large, it is expected that the time prediction unit 113 takes time for prediction, and the time for resource selection increases. A method for solving this is realized in the second embodiment.

  Next, the effect of this embodiment will be described. In the present embodiment, since the optimum state is calculated by predicting the state transition time, the resource addition can be performed in a relatively short time in response to the resource addition request. In Non-Patent Document 2, since the computer setting time is shortened by using a method that depends on NFS or the like, the computer setting time may not be shortened because it cannot be used depending on a service request. However, in this embodiment, since the computer setting time is shortened without using a specific method such as NFS, a versatile method for reducing the computer setting time is provided.

Embodiment 2. FIG.
Next, a second embodiment of the present invention will be described with reference to the drawings.

  In the present embodiment, for the purpose of reducing the time required for resource selection after receiving a service use request, the state determination unit stores information necessary for resource selection in advance.

  FIG. 8 is a block diagram showing a configuration of the management server 100 according to the second embodiment of the present invention. Also in the second embodiment, the management server 100 includes a data processing device 110 and a storage device 120. However, the data processing apparatus 110 according to the present embodiment also includes a use request reception unit 116 and a resource selection unit 117 in addition to the configuration of the first embodiment shown in FIG.

  Further, the operation of the state determination unit 112 is different from the operation in the first embodiment in that each resource stores the association of services that can change state in a short time. In the present embodiment, the state determination unit 112 stores information used when determining the state in the resource state management DB 124 as service association, and this information is used by the resource selection unit 117. As information on service association, for example, there is information including both information indicating a resource (computer resource) that can be used to provide a service and information indicating a set time of the resource.

  The use request receiving unit 116 receives a resource use request from a service user. The resource use request is that the service user designates the service to be used, the resource amount, and the attribute of the resource, and requests selection and setting of the resource to be used. The usage request reception unit 116 outputs the received resource usage request to the resource selection unit 117. At this time, the use request receiving unit 116 may perform user authentication, billing for the user, and the like.

  The resource selection unit 117 inputs a request from the use request reception unit 116, refers to information in the resource state management DB 124, and selects a resource corresponding to the request. The resource selection unit 117 uses the information stored in the resource state management DB 124 to select a resource with a short setting time from among resources that satisfy the service, resource amount, and resource attributes requested by the user. When a resource satisfying the request can be selected, a state transition request is made to the transition instruction unit 115. If a resource that satisfies the request cannot be selected, the user is informed that the resource could not be selected.

  The storage device 120 includes a resource state management DB 124 in addition to the configuration of the first embodiment shown in FIG. The resource state management DB 124 holds associations between services in which each resource can make a state transition in a short time, or associations between resources in which the transition time to a state where each service can be used, that is, a set time is short. By using this information by the resource selection unit 117, it is expected that the resource selection process executed by the resource selection unit 117 can be completed in a short time.

Next, the overall operation of the present embodiment will be described with reference to the flowcharts of FIGS. 9 and 10.
FIG. 9 shows a flow of processing in which processing necessary in the present embodiment is added to the state determination operation shown in FIG. The process from step A1 to A7 is the same as the process in the first embodiment. In the present embodiment, a resource state storage process is added after the process (resource state transition execution process) in which the state determination unit 112 in step A7 requests the transition instruction unit 115 to make a state transition (step A8). In the resource state storage process, the state determination unit 112 stores the association between a service and a resource having a short time for transitioning to a service usable state in the resource state management DB 124.

Next, with reference to FIG. 10, the resource use request acceptance process will be described.
First, the use request receiving unit 116 receives a resource use request from the service user (step E1). The resource use request includes information such as a service that the service user wants to use, a resource amount, and a resource attribute. Such information is used for resource selection. The amount of resources includes the number of computers, memory, CPU usage rate, and network bandwidth. The resource attribute is a physical attribute such as a computer resource architecture or network configuration, or a logical attribute such as a resource usage status, a service used, or a user. Further, it is assumed that the use request receiving unit 116 performs authentication of resource users, exchange of billing information, and the like. In the case of authenticating resource users, service provision to only specific users is realized. In addition, when charging information is exchanged, appropriate charging to the user is realized.

  Next, the resource selection unit 117 that has received the resource use request acquires information on the association between the service and the resource from the state determination unit 112 (step E2). This information is information generated by the state determination unit 112 in step A8 shown in FIG.

  Next, the resource selection unit 117 selects a resource that meets the request according to the information on the association between the resource and the service acquired at step E2 (step E3). At this time, the resource selection unit 117 acquires the resource state from the state acquisition unit 114 as necessary, or acquires the prediction time for the resource state transition from the transition time prediction unit 113. Based on this information, an optimum resource is selected in the required amount.

  Next, it is divided into two operations depending on whether or not the resource selection is successful (step E4). If the resource selection is successful, the resource selection unit 117 outputs a resource state transition execution request to the transition instruction unit 115. The transition instruction unit 115 issues an instruction to change the resource state to the agent 201 in the corresponding computer, and causes the resource state transition to be executed (step E5). Thereafter, the execution result of the resource state transition is received from the agent 201 and notified to the requesting user (step E6).

  If the resource selection has failed, the user who has requested the resource use request is notified that the resource selection has failed (step E7). The situation where resource selection fails is when there are no standby resources corresponding to the requested amount, or when requests such as resource attributes and resource setting time cannot be satisfied. At this time, the reason why the resource selection has failed may be notified. By determining the reason why the resource selection has failed, the user can change the request contents of the future resource use request so that the resource use request can be easily accepted.

  Next, the effect of this embodiment will be described. In the present embodiment, since the usage request reception unit 116 and the resource selection unit 117 from the user are included, it is possible to respond to the usage request from the user. Further, in the present embodiment, the state selection unit 112 can further shorten the resource selection time by executing a process of preliminarily storing resources for which the setting time of each service is short. That is, the state determination unit 112 realizes a resource selection unit that selects a computer resource to be used in advance (before the time when the resource use request is actually received).

Embodiment 3 FIG.
Next, a third embodiment of the present invention will be described with reference to the drawings.
The purpose of this embodiment is to simplify the input of state transition rules. It is complicated to manually generate a single state transition rule. Therefore, the procedure for setting each service or the state transition for setting the service is used as an input, and a single state transition rule is generated from the input, thereby simplifying the input of the state transition rule. .

  FIG. 11 is a block diagram illustrating a configuration of the management server 100 according to the third embodiment of this invention. Also in the third embodiment, the management server 100 includes a data processing device 110 and a storage device 120. However, the data processing apparatus 110 according to the present embodiment includes a transition procedure input unit 118 and a state transition synthesis unit 119 in addition to the configuration of the first embodiment shown in FIG. The state transition combining unit 119 is an example of a state transition rule unit that receives a plurality of state transition rules and combines them into one state transition rule.

  The transition procedure input unit 118 inputs a procedure (state transition procedure) for setting a computer resource in a state where a specific service can be used (state transition), and passes the procedure to the state transition synthesis unit 119. At this time, the transition procedure input unit 118 may perform authentication of the input person and confirmation of the procedure. The state transition synthesis unit 119 synthesizes the received state transition procedure into one state transition diagram. Specifically, multiple state transition procedures can be combined into a single state transition diagram by determining the same state and combining them into a single state, or by determining coexisting states and defining them as new states. To do. At this time, by reducing the number of states as much as possible, it is possible to shorten the time required for determining the optimum state. The synthesized state transition diagram is stored in the state transition rule DB 121 as a state transition rule.

  The operation of the present embodiment will be described in detail with reference to the flowchart of FIG. FIG. 12 shows operations from inputting a procedure for making a state transition (transition procedure) to generating and saving a state transition diagram.

  First, the transition procedure input unit 118 receives a plurality of transition procedures (step G1). At this time, the transition procedure input unit 118 may perform authentication of the input person and confirmation of the procedure. The transition procedure input unit 118 passes the input transition procedure to the state transition synthesis unit 119.

  Next, the state transition synthesizing unit 119 searches for the same state by comparing the states in the transition procedure. If there is the same state, the state transition synthesis unit 119 performs a process of collecting those procedures. By repeating the process, a plurality of transition procedures are combined into one state transition diagram (step G2). Finally, the state transition combining unit 119 stores the combined state transition diagram in the state transition rule DB 121 as a state transition rule.

  Next, the effect of this embodiment will be described. In the present embodiment, it is possible to determine an optimum state using a state transition rule by combining a plurality of transition procedures into one state transition rule. In addition, by comparing a plurality of states and reducing the number of states as much as possible, it is possible to reduce the time taken to determine the optimum state.

  Next, the operation of the embodiment of the present invention will be described using specific examples.

Example 1.
The first example corresponds to the first embodiment of the present invention. In this embodiment, the computer system includes one management server and a plurality of computers connected to the management server via an Ethernet (registered trademark) LAN (Local Area Network) as a network. Hereinafter, the management server having the configuration shown in FIG. 3 is taken as an example of the configuration of the management server.

  The state determination unit 112 that has received the processing start request from the start unit 111 acquires a state transition rule from the state transition rule DB 121.

  FIG. 13 is an explanatory diagram expressing the state transition rules in a state transition diagram. In the state transition diagram shown in FIG. 13, there are state 0 to state 9, state A, and state B, each of which is connected by an oriented branch. State 0 to state 9 are states in which the service is not operating, state A is a state in which service A can be provided (a state where the service can be used), and state B is a state in which service B can be provided. . Each state can transition with respect to a previous state connected by branches. For example, it is possible to transition from state 1 to state 2, but it is not possible to transition from state 2 to state 6. Transition starts from state 0, and transitions to state A or state B, which is a state in which a service can be provided via an intermediate state.

  Next, the state determination unit 112 acquires the state transition prediction time from the time prediction unit 113. The time prediction unit 113 acquires a transition time history from the state transition history DB 123. The history of transition time is held in the form shown in the explanatory diagram of FIG. In the example illustrated in FIG. 14, the history of transition from each state is retained in order. For example, the time required for the transition from the state 0 to the state 1 is 3 seconds (3 s) for the latest one, and 4 seconds for the previous one is exemplified.

  The time prediction unit 113 that has acquired the state transition history predicts the state transition time from the state transition history. For example, the average of the three latest transitions is used as the predicted value of the state transition time. In that case, the predicted value of the transition from state 1 to state 3 shown in FIG. 14 is 3 minutes (3 m). The time prediction unit 113 may pass the predicted value of the state transition time for each state transition to the state determination unit 112, or adds a predicted value to each branch of the state transition diagram as shown in the explanatory diagram of FIG. You may pass it in the form. FIG. 15 shows that, for example, the predicted time for transition from state 2 to state 5 is 14 minutes.

  The state determination unit 112 that acquired the predicted value of the transition time acquires a state determination index from the state determination index DB 122. FIG. 16 illustrates the arrival probability for each request when the state determination index is the arrival probability of the request. For example, it is shown that the arrival probability of a request for allocating one resource to service A is 0.3. This index is used by the state determination unit 112 to determine an optimal state.

  Next, the state determination unit 112 that has acquired the state determination index acquires the current resource state from the state acquisition unit 114. The state acquisition unit 114 determines in which state each resource is, and passes the determination result to the state determination unit 112. FIG. 17 illustrates an example in which the state of each resource is mapped to a state transition diagram. In FIG. 17, for example, resource 0 and resource 1 are shown to be in state 0, and resource 2 is shown to be in state 4.

  The state determination unit 112 that has obtained the information performs state determination processing. FIG. 18 and FIG. 19 are flowcharts showing a method for obtaining an optimum state by full search in the state determination process. FIG. 18 shows an operation of performing a full search, and FIG. 19 shows an operation of calculating an expected value of additional time in the full search. The additional time is the time from the state where the computer resource is placed until the setting for enabling the use of the service is completed. Therefore, the expected value of the additional time is, for example, a value obtained by adding the product of the request arrival probability and the predicted value of the additional time for each request when the state determination index is the request arrival probability.

  As shown in FIG. 18, when performing a full search, the state determination unit 112 first lists all combinations of states that each computer resource can take (step H1). Next, one combination is selected from the combination of states (step H2). And the expected value of additional time is calculated with respect to the state in the selected combination (step H3). It is confirmed whether there is any combination that is not selected (step H4), and step H3 is performed for all the combinations to calculate the expected value of the additional time. When the expected values have been calculated for all the combinations, the combination having the smallest expected value for the additional time is selected as the optimum state (step H5), and the process ends.

  FIG. 19 is a flowchart showing a procedure for calculating the expected value of the additional time in step H3 shown in FIG. As shown in FIG. 19, when calculating the expected value of the additional time, the state determination unit 112 first selects one service that can be used (step I1). Next, one computer resource (resource) is selected (step I2), and the minimum cost for state transition to the selected service is determined from the state of the selected computer resource (step I3). This is obtained by solving the shortest path problem from the state of the computer resource to the state where the service can be used. Since the shortest path problem is widely known, a description thereof is omitted here. The processing of step I3 is performed on all computer resources, and the minimum cost for transitioning from the state of all computer resources to the state where the service can be used is calculated.

  Next, n optimal computer resources for the service request are selected (step I5). The service arrival probability P is multiplied according to the number of individual computer resources to obtain a predicted expected value of the state transition time (step I6). The expected values of the selected combination can be obtained by performing the operations of Steps I2 to I6 for all services and requests and adding the predicted expected values (Step I8).

  When the optimum state is obtained as described above, each resource is transitioned to the optimum state by the transition instruction unit 115. For example, if resource 0 is currently in state 0 and the optimal state is determined to be state 5, the state of resource 0 is transitioned to state 5. These state transitions may be OS installation work, software installation, or software setting work. When the transition to the optimum state is completed for all resources, the entire process ends.

  The total search requires a large amount of calculation and is expected to take time to obtain an optimum value. In such a case, a state close to the optimum can be obtained in a short time by using a widely known optimization method, for example, an approximate solution such as a hill-climbing method.

Example 2
Next, a second embodiment will be described with reference to the drawings. The second example corresponds to the second embodiment. Also in this embodiment, the computer system is composed of one management server and a plurality of computers connected to the management server via an Ethernet LAN as a network. Hereinafter, the management server having the configuration shown in FIG. 8 is taken as an example of the configuration of the management server.

  In the operation for obtaining the optimum state, the difference between the first embodiment and the second embodiment is that, in the second embodiment, resources that have a short time to transition to a state where each service can be used. Is to execute the process of saving the association (process of step A8). In step I3 shown in FIG. 19 in the first embodiment, the minimum predicted time for performing a state transition from each computer resource to the service is obtained. In the present embodiment, the state determination unit 112 determines items to be stored in the resource state management DB 124 using the obtained minimum predicted time. For example, for each service, the state determination unit 112 stores, in the resource state management DB 124, information on a pair of each computer resource and the minimum prediction time for the computer resource in order of decreasing minimum prediction time.

  FIG. 20 shows an example of a table stored in the resource state management DB 124. In the example illustrated in FIG. 20, pairs of computer resources and predicted set times (predicted set time values) are stored in a state of being sorted using the predicted set time as a key corresponding to each service. FIG. 20 shows that, for example, the predicted value of the time required for the computer resource R0 to transition from the current state to the state where the service S0 can be provided is 3 seconds (3 s). Also, it is shown that the predicted value of the time required for the computer resource R3 to transition from the current state to the state where the service S0 can be provided is 7 seconds (7s). Further, it is shown that the predicted value of the time required for the computer resource R5 to transition from the current state to the state in which the service S0 can be provided is 10 hours (10h).

  Next, the resource use request receiving process will be described using the example shown in FIG. Assume that two resources are requested for the service S1 as a resource use request. At this time, the resource selection unit 117 performs selection processing by looking at the table held in the resource state management DB 124. In the table shown in FIG. 20, it is indicated that the setting time of the computer resource specified by R7 and the computer resource specified by R3 is short to add the service S1, so that the resource selection unit uses R7. And the computer resource specified by R3 are selected.

  The computer resource specified by the selected R7 and the computer resource specified by R3 are transitioned by the transition instruction unit 115 to the state of the service S1. If the transition is successful, the service user is informed that the transition is successful, and the process ends.

Example 3
Next, a third embodiment will be described with reference to the drawings. The third example corresponds to the third embodiment. Also in this embodiment, the computer system is composed of one management server and a plurality of computers connected to the management server via an Ethernet LAN as a network. Hereinafter, the management server having the configuration shown in FIG. 11 is taken as an example of the configuration of the management server.

  First, operations from receiving a state transition procedure to generating and storing a state transition diagram will be described. The transition procedure input unit 118 receives a plurality of state transition procedures. The state transition procedure is as shown in FIG. For example, the state transition procedure A indicates that it is necessary to repeat the transition from the state 0 to the state 1 and the state 2 in order to make the service A available (state A). The state transition procedure B2 indicates that the state B can be reached from the state 1 through the state 3 or through the state 4 in either case.

  The state transition synthesis unit 119 performs an operation of synthesizing these state transition procedures A, B1, and B2. From the state transition procedure A and the state transition procedure B1 shown in FIG. 21A, it can be determined that the transitions from the state 0 to the state 5 are the same. By judging such the same state, a state transition diagram as shown in FIG. 21B can be obtained. The state transition diagram obtained here is stored in the state transition rule DB 121 as a state transition rule.

  FIG. 22 shows a state transition diagram that has two different starting states (state 0 and state 10) and can return toward the starting state between the states. Even in such a state transition diagram, an optimum state can be obtained.

Example 4
Next, a fourth embodiment will be described.
Here, consider a case where a data center as a computer resource user borrows computer resources, for example, in units of time by utility computing, and provides a general user with a web-related service using the computer resources. Hereinafter, a data center as a computer resource user is referred to as a service provider. As service providers, portal site providers and online shopping services such as NEC (NEC) are conceivable.

  For example, when the number of accesses to a web page or the like is expected to increase, a service provider administrator issues a computer resource use request to the management server 100 via a terminal device or the like. The management server 100 operates as in each of the embodiments described above, and makes the computer resources available for providing services to the service provider. In this example, computer resources required by the service provider are made available to the service provider.

  Normally, a service provider borrows only the minimum computer resources necessary for providing services to general users. When the number of accesses is expected to increase, it is preferable to increase computer resources in a short time. Otherwise, the response speed to the access request to the web page or the like is lowered, and the quality of service for the general user by the service provider is lowered.

  As described above, according to each of the above-described embodiments and examples, since the computer resource is maintained in an optimal intermediate state, the service can be performed in a short time when requested by the service provider. The computer resource required by the provider can be used by the service provider. As a result, it is possible to prevent the quality of service provided to general users by the service provider from being degraded.

  The present invention can be applied to uses such as a state change device that determines and sets the state of a computer resource, and a program for realizing the state change device in a plurality of computer resources.

It is a block diagram which shows the computer system of 1st Embodiment. It is a block diagram which shows the structure of the computer for providing a service. It is a block diagram which shows the structure of a management server. It is a flowchart which shows the flow of the state determination operation | movement in 1st Embodiment. It is a flowchart which shows the flow of the transition time prediction operation | movement by the time estimation part in 1st Embodiment. It is a flowchart which shows the flow of the resource status acquisition operation | movement by the status acquisition part in 1st Embodiment. It is a flowchart which shows the flow of the state transition execution operation by the transition instruction | indication part in 1st Embodiment. It is a block diagram which shows the structure of the management server in 2nd Embodiment. It is a flowchart which shows the flow of the state determination operation | movement in 2nd Embodiment. It is a flowchart which shows the resource utilization request reception process in 2nd Embodiment. It is a block diagram which shows the structure of the management server in 3rd Embodiment. It is a flowchart which shows the flow of operation | movement from inputting a transition procedure to producing | generating and storing a state transition diagram. It is explanatory drawing which expressed the state transition rule in 1st Example with the state transition diagram. It is explanatory drawing which shows the content of the transition time log | history DB in a 1st Example. It is explanatory drawing which shows the state transition diagram to which the state transition prediction time in the 1st Example was added. It is explanatory drawing which shows the arrival probability with respect to each request | requirement in case the state determination parameter | index in a 1st Example is the arrival probability of a request | requirement. It is explanatory drawing which shows the state transition diagram to which the resource state and state transition prediction time were added in the 1st Example. It is a flowchart which shows the operation | movement which performs the resource state determination process in a 1st Example by a full search. It is a flowchart which shows the operation | movement which calculates | requires the expected value of the additional time of the resource in a 1st Example. It is explanatory drawing which shows the example of the table preserve | saved in resource state management DB in a 2nd Example. It is explanatory drawing which shows the state transition diagram synthesize | combined from the state transition procedure in a 3rd Example. It is explanatory drawing which shows the state transition diagram in which it is considered that it has a some start state in 3rd Example, and each state returns.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Management server 101 Network 102, 103, 104 Computer 110 Data processing apparatus 111 Start part 112 State determination part 113 Time prediction part 114 State acquisition part 115 Transition instruction part 116 Usage request reception part 117 Resource selection part 118 Transition procedure input part 119 Transition State composition unit 120 Storage device 121 State transition rule DB
122 State determination index DB
123 Transition time history DB
124 Resource state management DB

Claims (19)

In a computer system having a plurality of computer resources,
And a state determination means for determining a state of the computer resource in which an expected value of the set time is reduced by using a predicted value of the set time that is a time required for setting the computer resource and a setting procedure of the computer resource. A computer system.   2. The computer system according to claim 1, further comprising state transition means for presetting computer resources to a state determined by the state determination means. State transition time prediction means for predicting the set time;
A state transition rule storage means for storing a state transition rule representing a computer resource setting procedure,
The computer system according to claim 1, wherein the state determination unit uses a predicted value predicted by the state transition time prediction unit and a setting procedure stored in the state transition rule storage unit. 4. The computer system according to claim 3, further comprising state transition rule means for receiving a plurality of state transition rules and combining them into one state transition rule. The computer system according to any one of claims 1 to 4, further comprising resource selection means for selecting a computer resource to be used in advance. Comprising a state determination index storage means for storing an index for determining the state of a computer resource;
The computer system according to any one of claims 1 to 5, wherein the state determination unit determines a state of a computer resource using an index stored in the state determination index storage unit. A management server for managing each computer in a computer system having a plurality of computer resources,
A state determination means for determining a state of the computer resource that minimizes the expected value of the set time by using the predicted value of the set time, which is the time required for setting the computer resource, and the setting procedure of the computer resource. A management server characterized by that.   2. The management server according to claim 1, further comprising state transition means for pre-setting computer resources to a state determined by the state determination means. A method for reducing computer setting time in a computer system having a plurality of computer resources,
Decrease the computer setup time, which is characterized by determining the state of the computer resource where the expected value of the computer resource setup time is reduced using the predicted value of the setup time, which is the time taken to set up the computer resource, and the setup procedure Method. The method for reducing the computer setting time according to claim 9, wherein the computer resource is preset in a state determined by the state determination means. The method for reducing the computer setting time according to claim 9 or 10, wherein a computer resource to be used is selected in advance. Perform a prediction process to predict the set time,
The state value of the prediction value by the said prediction process and the state transition rule showing the setting procedure of the computer resource memorize | stored beforehand are used when determining the state of a computer resource. A method for reducing the computer setting time described in 1. The method for reducing computer resource setting time according to claim 12, wherein a plurality of state transition rules are received and combined into one state transition rule. The method for reducing the computer setting time according to any one of claims 9 to 13, wherein an index for determining the state of the computer resource is used when determining the state of the computer resource. In a computer that manages each computer in a computer system having a plurality of computer resources,
Computer setting to execute the state determination process that determines the state of the computer resource where the expected value of the setting time becomes smaller using the predicted value of the setting time, which is the time required to set the computer resource, and the setting procedure of the computer resource A program to reduce time. On the computer,
The program for reducing a computer setting time according to claim 15, for executing a state transition process for presetting a state of a computer resource to a state determined by the state determination process. On the computer,
A process of storing a state transition rule representing a computer resource setting procedure in the state transition rule storage means;
The state transition time prediction process for predicting the set time is executed,
The computer setting according to claim 15 or 16, for executing the state determination process using the predicted value predicted in the state transition time prediction process and the setting procedure stored in the state transition rule storage unit. A program to reduce time. On the computer,
The program for reducing the computer resource setting time according to claim 17 for executing a state transition rule process for receiving a plurality of state transition rules and combining them into one state transition rule. On the computer,
Execute the process to memorize the index when determining the state of computer resources,
The program for reducing a computer resource setting time according to any one of claims 15 to 18, for executing a state determination process using an index stored in the state determination index storage means. .

Images