JP5978673B2 - Information processing device - Google Patents

Description

  The present invention relates to an information processing device, an allocation device, an information processing method, an allocation method, an information processing program, and an allocation program.

  Conventionally, there is a technique for reducing the number of servers operating in a server room by collecting a plurality of VMs (Virtual Machines) assigned to servers in the server room and reassigning them to one processor of any server (For example, refer to Patent Document 1 below.)

JP 2011-8822 A

  However, in the above-described conventional technology, if the processor is in a state in which a plurality of VMs can be executed, the processor executes a plurality of VMs without considering the type of VM. For this reason, the performance of the VM may deteriorate due to competition between the plurality of VMs. When the VM performance deteriorates, the response performance defined in SLA (Service Level Agreement) may not be ensured due to a contract between the VM provider and the VM user. As a result, the response performance of the VM may be reduced in transactions such as on the network, and the transaction may fail. On the other hand, a VM provider may have to compensate a VM user for an SLA violation.

  An object of this invention is to suppress the performance fall of VM.

  According to one aspect of the present invention, a first type virtual machine that executes predetermined processing by bidirectional communication with another device, and a fixed form that does not have interactive input / output and performs predetermined processing by a predetermined time A virtual machine group that includes a second type of virtual machine that executes processing, and a virtual machine group in which a predicted value of the arithmetic processing unit usage rate of the virtual machine group is equal to or less than a threshold is received from the allocating device, The destination is determined to be one of the plurality of arithmetic processing devices so as to be within the processing capability of the arithmetic processing device, and the virtual processing unit determined as the placement destination among the plurality of arithmetic processing devices is virtually An information processing apparatus, an information processing method, and an information processing program for assigning a machine group and controlling a first type virtual machine to be executed with priority over a second type virtual machine are proposed.

  Further, according to one aspect of the present invention, there is no interactive input / output with a first type of virtual machine that executes predetermined processing by bidirectional communication with another device from the server group. Processing of the virtual machine group when selecting the allocation destination of the virtual machine group including the second type of virtual machine that executes the standard processing that is performed by a predetermined time and allocating the virtual machine group to the same server Obtain a predicted value of the processing device usage rate based on the device usage rate, determine whether the acquired predicted value is less than or equal to a threshold value based on the number of arithmetic processing devices of the selected server, and threshold An allocation apparatus, an allocation method, and an allocation program for allocating a virtual machine group to a selected server so that the first type of virtual machine is executed in preference to the second type of virtual machine when it is determined to be less than or equal to the value Is proposed.

  According to one aspect of the present invention, there is an effect that it is possible to suppress a decrease in VM performance.

FIG. 1 is an explanatory diagram showing the contents of VM allocation to an information processing apparatus by the allocation apparatus. FIG. 2 is an explanatory diagram showing a system configuration example. FIG. 3 is a block diagram illustrating a hardware configuration example of a computer. FIG. 4 is an explanatory diagram showing an example of the stored contents of the server room structure information. FIG. 5 is an explanatory diagram of an example of the contents stored in the VM number information. FIG. 6 is an explanatory diagram of an example of the stored contents of the maximum executable VM number table. FIG. 7 is an explanatory diagram of an example of the contents stored in the VM power consumption table. FIG. 8 is an explanatory diagram of an example of the contents stored in the base power consumption table. FIG. 9 is an explanatory diagram showing an example of the stored contents of scaling information. FIG. 10 is an explanatory diagram of an example of the contents stored in the VM arrangement table. FIG. 11 is an explanatory diagram of an example of the contents stored in the power supply table. FIG. 12 is an explanatory diagram of an example of the contents stored in the rack state table. FIG. 13 is an explanatory diagram of an example of the contents stored in the initial arrangement table. FIG. 14 is an explanatory diagram showing an example of the contents stored in the UPS table. FIG. 15 is an explanatory diagram of an example of the contents stored in the allocation target VM table. FIG. 16 is an explanatory diagram of an example of the stored contents of the arranged VM table. FIG. 17 is a block diagram illustrating a functional configuration example of the control server 100. FIG. 18 is a block diagram illustrating a functional configuration example of the server S. FIG. 19 is an explanatory diagram (part 1) illustrating an example of a VM allocation result to the server S by the control server 100. FIG. 20 is an explanatory diagram (part 2) illustrating an example of a VM allocation result to the server S by the control server 100. FIG. 21 is a flowchart showing a detailed processing procedure of the allocation processing.

  Exemplary embodiments of an information processing device, an allocation device, an information processing method, an allocation method, an information processing program, and an allocation program according to the present invention will be described below in detail with reference to the accompanying drawings. In the following, a server in a server room is employed as the information processing apparatus, and a control server that allocates VMs to the server is employed as the allocation apparatus.

  Here, the VM allocated to the server includes, for example, a VM for securities trading that is interactive with the client, and a VM for collecting sales data that is not interactive with the client. Since the CPU usage rate of the interactive VM depends on the communication frequency and communication timing from the client, the fluctuation of the peak of the CPU usage rate is large and the timing of the fluctuation is random. When such different types of VMs are assigned to the same server, for example, even when the CPU usage rate of the interactive VM is high, the VM for sales data aggregation uses a certain CPU resource, so that the server resources are reduced. Tighten.

  As a result, the performance of the VM for securities trading deteriorates, and the securities trading may fail due to, for example, being unable to communicate with the client. Also, if the CPU resource of a stock transaction VM continues to be used at a high CPU utilization rate, the VM for sales data aggregation will not be completed by the specified end time, and the client will not be able to use the aggregation results, which is disadvantageous May suffer.

  The control server according to the present embodiment assigns each VM to a server that satisfies both the condition that the performance of the VM for stock trading does not deteriorate and the condition that the VM for sales data aggregation ends by the specified end time. assign. Then, the server executes each assigned VM by one CPU. In this case, the server executes the stock transaction VM with priority over the VM for sales data aggregation.

  As a result, the server can efficiently use CPU resources. For example, when a VM for securities trading is heavily loaded, the server causes the VM for securities trading to use the CPU resource used by the VM for summarizing sales data, thereby reducing the performance of the VM for securities trading. Can run without. Further, since the server satisfies the condition that the VM for sales data aggregation ends by the specified end time, even if the VM for securities trading is executed preferentially, the VM for sales data aggregation is set to the specified end time. Can be finished by

  Further, as a result of assigning a plurality of VMs to one server, if there is a server to which no VM is assigned, the power supply to the server can be stopped and the power consumption of the server room can be reduced. Accordingly, the number of servers in operation in the server room is reduced, and the power consumption of the server room is reduced.

  FIG. 1 is an explanatory diagram showing the contents of VM allocation to an information processing apparatus by the allocation apparatus. In FIG. 1, the control server 100 is a computer that assigns VMs to the server S so that the number of servers S to which VMs are assigned decreases. The VM includes a VM that executes a predetermined process through two-way communication with another device (for example, a client), and a VM that executes a standard process with a specified end time set.

  Hereinafter, a VM that executes a predetermined process by bidirectional communication with another device is referred to as an “online VM”. An example of the online VM is a VM that performs the above-described securities transaction. Further, a VM that executes the standard process is referred to as a “batch VM”. The batch VM includes, for example, a VM that aggregates the sales data described above.

  Here, since the online VM performs processing by bidirectional communication with the client, when the performance of the online VM deteriorates, processing for a request from the client and response to the client may be delayed. As a result, client users may suffer disadvantages or SLA violations may occur. For example, when the performance of a VM that conducts securities transactions deteriorates, there are cases where the client user purchases or sells securities and the client users are damaged.

  In addition, since the specified end time is set for the batch VM, if the user does not end by the specified end time, the user of the processing result of the batch VM may suffer a disadvantage or cause an SLA violation. For example, in a VM that aggregates sales data, if the aggregation of sales data used to start business on the next business day does not end, the user cannot start business and damages the user.

  Therefore, when allocating a VM to the server S, the control server 100 avoids a decrease in the performance of the online VM, the batch VM ends by the specified end time, and the number of servers S to which the VM is allocated decreases. To. Specifically, for example, the control server 100 assigns VMs one by one to the server S that satisfies the conditions of the following formulas (1) and (2).

The above equation (1) indicates a condition for ending the batch VM by the specified end time. Here, the average CPU usage rate of the batch VM is the minimum average CPU usage rate at which the batch VM can be completed by the specified end time. The average CPU usage rate of the online VM is an average value calculated from the measured CPU usage rate of the online VM. The average CPU usage rate is stored in the control server 100. The P _i, a CPU utilization rate upper limit of the server S, a value obtained by multiplying the upper limit of the CPU utilization of the number and 1CPU per CPU in the server S.

  Specifically, the condition of the above formula (1) is the minimum average that can finish each of the plurality of batch VMs by the specified end time even when the average CPU utilization of the plurality of online VMs is taken into consideration. It shows that the CPU usage rate can be secured below the CPU usage rate upper limit of the server S. Therefore, the server S that satisfies the condition of the above formula (1) can finish the batch VM by the specified end time.

  The above equation (2) represents a condition for avoiding the performance degradation of the online VM. Here, σ is a standard deviation in the normal distribution of the CPU usage rate of the online VM. c is a multiple of the standard deviation. For example, a natural number of c = 1 to 3 is adopted as c. Therefore, “average VM usage rate of online VM + σ of CPU usage rate of online VM” × c is the CPU usage rate when the online VM is heavily loaded.

  Specifically, the condition of the above formula (2) is that the sum of the CPU utilization rates of the online VMs is equal to or less than the upper limit of the CPU utilization rate of the server S even when a plurality of online VMs are simultaneously loaded. It shows that it becomes. Therefore, the server S that satisfies the condition of the above formula (2) can avoid competition between online VMs, and can avoid performance degradation of online VMs.

  The server S is a computer that executes an assigned VM by any of the CPUs in the own device. Further, when there is an online VM among the allocated VMs, the server S executes the online VM with priority. For example, the server S uses CPU resources preferentially for execution of online VMs, and uses excess CPU resources not used for execution of online VMs for execution of batch VMs. Thereby, the server S can avoid the performance degradation of online VM. In FIG. 1, the server S has three CPUs.

  In the example of FIG. 1, the control server 100 assigns the online VM 1, online VM 2, batch VM 3, and batch VM 4 to the server S. First, the control server 100 selects the online VM 1 as a VM to be allocated. Here, it is assumed that the average CPU usage rate of the online VM 1 is “30%”, the standard deviation σ of the CPU usage rate of the online VM 1 is “20%”, and the constant c is “3”. Here, the CPU utilization rate upper limit Pi of the servers S1 and S2 is set to a value “240%” obtained by multiplying the upper limit “80%” of the CPU utilization rate per CPU by the number of CPUs “3”.

  The control server 100 selects the server S1 as the server S to be assigned. When the online server VM1 is assigned to the selected server S1, the control server 100 determines whether or not the conditions of the above formulas (1) and (2) are satisfied in the server S1. The condition of the above formula (1) is “30% ≦ 240%”, which satisfies the condition. The condition of the above formula (2) is “30% + 20% × 3 = 90% ≦ 240%”, which satisfies the condition. Therefore, the control server 100 allocates the online VM1 to the server S1.

  Next, the control server 100 selects the online VM 2 as the allocation target VM. Here, the average CPU usage rate of the online VM 2 is “30%”, the standard deviation σ of the CPU usage rate of the online VM 2 is “20%”, and the constant c is “3”. The control server 100 selects the server S1 to which the online VM1 has been allocated as the server S that is the allocation destination so that the number of servers S to which the VM is allocated decreases. When the online VM 2 is assigned to the selected server S1, the control server 100 determines whether or not the conditions of the above formulas (1) and (2) are satisfied in the server S1.

  Here, the equation (1) satisfies “30% + 30% = 60% ≦ 240%”, and therefore satisfies the condition of the equation (1). Further, since the above equation (2) satisfies “30% + 20% × 3 + 30% + 20% × 3 = 180% ≦ 240%”, the condition of the equation (2) is satisfied. Therefore, the control server 100 allocates the online VM2 to the server S1.

  Next, the control server 100 selects the batch VM 3 as a VM to be assigned. Here, the average CPU utilization of the batch VM 3 is “30%”. The control server 100 selects the server S1 to which the online VM1 and the online VM2 have been allocated as the allocation destination server S so that the number of servers S to which the VM is allocated decreases. When the batch server VM3 is assigned to the selected server S1, the control server 100 determines whether or not the conditions of the above formulas (1) and (2) are satisfied in the server S1.

  Here, the equation (1) satisfies “30% + 30% + 30% = 90% ≦ 240%”, and therefore satisfies the condition of the equation (1). Further, since the above equation (2) satisfies “30% + 20% × 3 + 30% + 20% × 3 = 180% ≦ 240%”, the condition of the equation (2) is satisfied. Therefore, the control server 100 assigns the batch VM3 to the server S1.

  Next, the control server 100 selects the batch VM 4 as the VM to be assigned. Here, the average CPU utilization of the batch VM 4 is “30%”. The control server 100 selects the server S1 to which the online VM1, the online VM2, and the batch VM3 are allocated as the server S that is the allocation destination so that the number of servers S to which the VM is allocated decreases. When the batch server VM4 is assigned to the selected server S1, the control server 100 determines whether or not the conditions of the above formulas (1) and (2) are satisfied in the server S1.

  Here, the equation (1) satisfies “30% + 30% + 30% + 30% = 120% ≦ 240%”, and therefore satisfies the condition of the equation (1). Further, since the above equation (2) satisfies “30% + 20% × 3 + 30% + 20% × 3 = 180% ≦ 240%”, the condition of the equation (2) is satisfied. Therefore, the control server 100 assigns the batch VM4 to the server S1. The control server 100 selects a server S different from the selected server S and reselects the different server when the server S that is the selected allocation destination does not satisfy the condition of the above formula (1) or formula (2). A VM is assigned to S.

  The server S1 executes the assigned online VM1, online VM2, batch VM3, and batch VM4 by any of the CPUs in the own device. Here, the server S1 executes with higher priority of the online VM1 and the online VM2. Further, the server S1 may cause each of the CPU 101 and the CPU 102 to execute one online VM so that the online VM1 and the online VM2 do not compete with each other.

  As shown in FIG. 1, the server S1 executes an online VM1, a batch VM3, and a batch VM4 by the CPU 101. The server S1 executes the online VM2 by the CPU 102.

  Here, for the batch VM3 and the batch VM4, the server S1 executes processing using CPU resources that are not used by the online VM1, and ends the processing without causing the performance degradation of the online VM1. The server S1 executes the online VM1 and the online VM2 with priority over the batch VM3 and the batch VM4. For this reason, the online VM1 and the online VM2 use the same CPU resources as when executed alone, and can ensure the same performance as when executed alone.

  As described above, the control server 100 allocates a plurality of VMs to one server S as long as the conditions of the above formulas (1) and (2) are satisfied. As a result, the control server 100 can avoid the performance degradation of the online VM, allocate the VM so that the batch VM is completed by the specified end time, and the number of servers S to which the VM is allocated decreases. it can.

  If there is a server S to which no VM is assigned, the control server 100 stops power supply to the server S. Thereby, the power consumption of a server room is reduced. Further, the control server 100 allocates a plurality of VMs to one server S, so that the server S can efficiently use CPU resources.

Further, the control server 100 can allocate the online VM1, online VM2, batch VM3, and batch VM4 to the server S by solving the mixed integer programming problem with the above formulas (1) and (2) as constraints. Good. For example, using a mixed integer programming problem that takes as input the average CPU utilization rate of a deployed VM and the standard deviation of the CPU utilization rate and outputs a variable x _{n, i, j} , a variable v _ij, and a variable θ _i , The VM is assigned to the server S.

The variable x _{n, i, j} indicates whether VM _n is assigned to the server S (i, j). The variable v _ij indicates whether the power supply of the server S (i, j) is on or off. The variable θ _i indicates whether a VM is assigned to the rack (i). i represents a rack number, and j represents a server number in the i-th rack. n indicates the number of the VM to be allocated. A specific example of assigning a VM to the server S using the mixed integer programming problem will be described later with reference to FIG.

Moreover, said Formula (1) and Formula (2) are conditions in case the performance of the server S is the same. When the performance of the server S is different, the following formula (3) is used instead of the above formula (1), and the following formula (4) is used instead of the above formula (2). Here, γ _i is a proportionality coefficient indicating how many times the performance of the server S is higher than that of the reference server S. Γ _i is acquired in advance by a benchmark test for the server S. Further, the control server 100 may update the γ _i by executing a benchmark test on the server S at regular time intervals.

  In addition, the control server 100 may select a server S with high cooling efficiency by CRAC (Computer Room Air Conditioning) in the server room as the server S to which the VM is allocated. Thereby, the power consumption of CRAC is reduced.

(System configuration example)
Next, a system configuration example will be described with reference to FIG.

  FIG. 2 is an explanatory diagram showing a system configuration example. In FIG. 2, the system 200 includes a server room, a control server 100, a task input server 201, and a client 202. The server room includes a rack 203 having a plurality of servers S, a CRAC 204, and a chiller 205 (outdoor unit).

  The server S group in the server room, the control server 100, the task input server 201, and the client 202 are connected by a network NET. The network NET is composed of, for example, a LAN (Local Area Network), a WAN (Wide Area Network), the Internet, a mobile phone network, and the like.

  The task input server 201 is a computer that transmits an allocation request to the server S of the VM to the control server 100. The allocation request includes, for example, a VM execution file and information indicating whether the VM is an online VM or a batch VM.

  The control server 100 is a computer that receives an allocation request from the task input server 201 to the VM server S. The control server 100 avoids the performance degradation of the online VM, the received VM is a server in the server room so that the batch VM is finished by the specified end time, and the number of servers S to which the VM is assigned is reduced. Assign to S.

  The client 202 is a computer that performs bidirectional communication with the online VM. The client 202 transmits a processing request to the online VM and receives a processing result. The client 202 may be a computer that receives the processing result of the batch VM. As the client 202, for example, a desktop PC (Personal Computer), a notebook PC, a mobile phone (smart phone, PHS (Personal Handyphone System)), or a tablet terminal can be employed.

  The rack 203 is equipped with S servers S and UPS (Uninterruptable Power Supply) 206. For example, assuming that there are N racks 203 in the server room, the i-th rack 203 is represented as a rack 203 (i). The server S is a computer that executes the assigned VM. When there is an online VM among the allocated VMs, the server S preferentially executes the online VM. For example, the j-th server S of the i-th rack 203 is represented as a server S (i, j). The UPS 206 is a device that manages power supply to the server S. For example, the UPS 206 of the i-th rack 203 is represented as UPS 206 (i).

  The CRAC 204 is a device that cools the server S in the rack 203. For example, assuming that CRACs 204 exist in the server room, the kth CRAC 204 is represented as CRAC 204 (k). The chiller 205 is an apparatus that exhausts air. Assume that the same number of chillers 205 as the CRACs 204 exist, and for example, the k-th chiller 205 is represented as chiller 205 (k).

(Computer hardware configuration example)
Next, a hardware configuration example of a computer employed as the control server 100 and the server S will be described with reference to FIG.

  FIG. 3 is a block diagram illustrating a hardware configuration example of a computer. In FIG. 3, a computer 300 is configured by connecting CPUs 301, a storage device 302, an input device 303, an output device 304, and a communication device 305 to a bus 310.

  The CPUs 301 govern the overall control of the computer 300. In addition, the CPUs 301 execute various programs stored in the storage device 302 to read data in the storage device 302 and write data as an execution result to the storage device 302. The various programs include an OS (Operating System) and the information processing program or assignment program of the present embodiment.

  The storage device 302 includes a ROM (Read Only Memory), a RAM (Random Access Memory), a flash memory, a magnetic disk drive, and the like. The storage device 302 serves as a work area for the CPUs 301, and various programs and data obtained by executing each program. Or store various data including

  The input device 303 is an interface for inputting various data by a user operation such as a keyboard, a mouse, and a touch panel. The output device 304 is an interface that outputs data according to an instruction from the CPUs 301. For example, a display or a printer is employed as the output device 304. The communication device 305 is an interface that receives data from the outside via the network NET and transmits data to the outside.

(Storage contents of server room structure information)
Next, an example of the contents stored in the server room structure information will be described with reference to FIG. The server room structure information is information indicating the number of racks and the number of servers in the server room. The server room structure information is information referred to when solving the mixed integer programming problem. The server room structure information is realized by the storage device 302, for example.

  FIG. 4 is an explanatory diagram showing an example of the stored contents of the server room structure information. As illustrated in FIG. 4, the server room structure information 400 includes a rack number item and a server number item. The number of racks 203 existing in the server room is stored in the rack number item. In the server number field, the number of servers S existing in one rack 203 is stored. The value of each item is previously input by the user of the control server 100, for example.

(Storage contents of VM number information)
Next, an example of the contents stored in the VM number information will be described with reference to FIG. The VM number information is information referred to when solving the mixed integer programming problem. The VM number information is realized by the storage device 302, for example.

  FIG. 5 is an explanatory diagram of an example of the contents stored in the VM number information. As illustrated in FIG. 5, the VM number information 500 includes a number item. The number field stores the number of VMs to be allocated. For example, when the allocation target VM increases, the control server 100 adds the number of the increase to the number item. For example, when the VM to be allocated is allocated to the server S, the control server 100 subtracts the allocated number from the number item.

(Storage contents of the maximum executable VM count table)
Next, an example of the stored contents of the maximum executable VM number table will be described with reference to FIG. The maximum executable VM number table is a table that stores the maximum executable VM number for each server. The maximum executable VM number table is referred to when solving the mixed integer programming problem. The maximum executable VM number table is realized by the storage device 302, for example.

  FIG. 6 is an explanatory diagram of an example of the stored contents of the maximum executable VM number table. As illustrated in FIG. 6, the executable maximum VM number table 600 includes a rack number item, a server number item, and a maximum executable VM number item, and configures a record for each server S.

  In the rack number item, a number for identifying the rack 203 existing in the server room is stored. The server number item stores a number for identifying the server S in the rack 203 indicated by the rack number item. The maximum executable VM number item stores the maximum number of VMs that can be executed by the server S indicated in the server number item. The value of each item is previously input by the user of the control server 100, for example.

(Storage contents of VM power consumption table)
Next, an example of the contents stored in the VM power consumption table will be described with reference to FIG. The VM power consumption table is a table that stores power consumption per VM for each server. The VM power consumption table is referenced when solving the mixed integer programming problem. The VM power consumption table is realized by the storage device 302, for example.

  FIG. 7 is an explanatory diagram of an example of the contents stored in the VM power consumption table. As illustrated in FIG. 7, the VM power consumption table 700 includes a rack number item, a server number item, and a VM power consumption item, and configures a record for each server S.

  In the rack number item, a number for identifying the rack 203 existing in the server room is stored. The server number item stores a number for identifying the server S in the rack 203 indicated by the rack number item. The VM power consumption item stores the power consumed to execute one VM on the server S indicated by the server number item. The value of each item is previously input by the user of the control server 100, for example.

(Storage contents of base power consumption table)
Next, an example of the contents stored in the base power consumption table will be described with reference to FIG. The base power consumption table is a table that stores constant power consumption for each server S. The base power consumption table is referred to when solving the mixed integer programming problem. The base power consumption table is realized by the storage device 302, for example.

  FIG. 8 is an explanatory diagram of an example of the contents stored in the base power consumption table. As illustrated in FIG. 8, the base power consumption table 800 includes a rack number item, a server number item, and a base power consumption item, and configures a record for each server S.

  In the rack number item, a number for identifying the rack 203 existing in the server room is stored. The server number item stores a number for identifying the server S in the rack 203 indicated by the rack number item. The base power consumption item stores the constant power consumption of the server S indicated by the server number item. The constant power consumption is power consumed when the server S is not executing processing. The value of each item is previously input by the user of the control server 100, for example.

(Storage information of scaling information)
Next, an example of storage contents of scaling information will be described with reference to FIG. The scaling information is information indicating a constant to be multiplied with respect to the standard deviation. The scaling information is information that is referred to when solving the mixed integer programming problem. The scaling information is realized by the storage device 302, for example.

  FIG. 9 is an explanatory diagram showing an example of the stored contents of scaling information. As shown in FIG. 9, the scaling information 900 includes constant items. In the constant item, a constant c used for calculating the upper limit of the CPU usage rate of the online VM is stored. The constant c is a multiple of the standard deviation σ. For example, a natural number of c = 1 to 3 is adopted as c. The constant c is given by the above formula (2), for example. The value of c is input by the user of the control server 100 in advance, for example.

(Storage contents of VM placement table)
Next, an example of the contents stored in the VM arrangement table will be described with reference to FIG. The VM allocation table is a table that stores information indicating whether a VM is allocated to each server. The VM allocation table stores a solution of the mixed integer programming problem. The VM arrangement table is realized by the storage device 302, for example.

  FIG. 10 is an explanatory diagram of an example of the contents stored in the VM arrangement table. As illustrated in FIG. 10, the VM arrangement table 1000 includes a VM number item, a rack number item, a server number item, and an arrangement destination item, and configures a record for each server S.

  In the VM number item, a number for identifying the VM is stored. In the rack number item, a number for identifying the rack 203 existing in the server room is stored. The server number item stores a number for identifying the server S in the rack 203 indicated by the rack number item. The placement destination item stores information indicating whether or not the VM indicated by the VM number item is allocated to the server S indicated by the server number item. Here, “0” in the placement destination item indicates that a VM is not allocated, and “1” indicates that a VM is allocated.

(Memory table contents)
Next, an example of the contents stored in the power supply table will be described with reference to FIG. The power table is a table that stores information indicating whether power is supplied to each server. The power table is referenced when solving the mixed integer programming problem. The power supply table is realized by the storage device 302, for example.

  FIG. 11 is an explanatory diagram of an example of the contents stored in the power supply table. As illustrated in FIG. 11, the power supply table 1100 includes a rack number item, a server number item, and a power supply item, and configures a record for each server S.

  In the rack number item, a number for identifying the rack 203 existing in the server room is stored. The server number item stores a number for identifying the server S in the rack 203 indicated by the rack number item. The power item stores whether or not power is supplied to the server S indicated by the server number item. Here, the power item “0” indicates that no power is supplied, and “1” indicates that the power is supplied. The value of the power item is updated by acquiring information indicating whether power is supplied to the server S from the UPS 206, for example.

(Contents stored in the rack status table)
Next, an example of the contents stored in the rack state table will be described with reference to FIG. The rack state table is a table that stores information indicating whether a VM is allocated to each rack. The rack state table is referred to when solving the mixed integer programming problem. The rack state table is realized by the storage device 302, for example.

  FIG. 12 is an explanatory diagram of an example of the contents stored in the rack state table. As illustrated in FIG. 12, the rack state table 1200 includes a rack number item and a VM execution item, and configures a record for each rack 203. In the rack number item, a number for identifying the rack 203 existing in the server room is stored. The VM execution item stores information indicating whether a VM is being executed on the rack 203 indicated by the rack number item. For example, “0” of the VM execution item indicates that the VM is not executed, and “1” indicates that the VM is executed. For example, the control server 100 updates the VM execution item every time a VM is allocated. For example, when the server S is notified that the execution of the VM has ended, the control server 100 updates the VM execution item.

(Storage contents of the initial allocation table)
Next, an example of the contents stored in the initial arrangement table will be described with reference to FIG. The initial placement table is a table that stores the number of placed VMs for each server. The initial allocation table is referred to when solving the mixed integer programming problem. The initial arrangement table is realized by the storage device 302, for example.

  FIG. 13 is an explanatory diagram of an example of the contents stored in the initial arrangement table. As illustrated in FIG. 13, the initial arrangement table 1300 includes a rack number item, a server number item, and an initial arrangement item, and configures a record for each server S. In the rack number item, a number for identifying the rack 203 existing in the server room is stored. The server number item stores a number for identifying the server S in the rack 203 indicated by the rack number item. The initial arrangement item stores the number of VMs arranged in the server S indicated by the server S item.

(Storage contents of UPS table)
Next, an example of the contents stored in the UPS table will be described with reference to FIG. The UPS table is a table that stores the constant power consumption and the proportionality coefficient by which the power consumption of the server is multiplied for each UPS. The UPS table is referred to when solving the mixed integer programming problem. The UPS table is realized by the storage device 302, for example.

  FIG. 14 is an explanatory diagram showing an example of the contents stored in the UPS table. As illustrated in FIG. 14, the UPS table 1400 includes a UPS number item, a base power consumption item, and a proportional coefficient item, and configures a record for each UPS 206. In the UPS number item, a number for identifying the UPS 206 existing in the server room is stored. The base power consumption item stores the constant power consumption of the UPS 206. In the proportional coefficient item, a coefficient indicating how much power is consumed by the UPS 206 with respect to the power consumption of the server S is stored. The value of each item is previously input by the user of the control server 100, for example.

(Storage contents of the allocation target VM table)
Next, an example of the contents stored in the allocation target VM table will be described with reference to FIG. The allocation target VM table is a table that stores an average CPU usage rate and a standard deviation of the CPU usage rate for each VM to be allocated. The allocation target VM table is referred to when the mixed integer programming problem is solved. The allocation target VM table is realized by the storage device 302, for example.

  FIG. 15 is an explanatory diagram of an example of the contents stored in the allocation target VM table. As illustrated in FIG. 15, the allocation target VM table 1500 includes a VM number item, an average CPU usage rate item, and a standard deviation item, and configures a record for each VM to be allocated. In the VM number field, a number for identifying a VM to be allocated is stored. The average CPU usage rate item stores the average CPU usage rate of the VM indicated by the VM number item. The standard deviation item stores the standard deviation of the VM CPU utilization indicated by the VM number item.

(Storage contents of the placed VM table)
Next, an example of the stored contents of the arranged VM table will be described with reference to FIG. The placed VM table is a table that stores an average CPU usage rate and a standard deviation of the CPU usage rate for each placed VM. The placed VM table is referred to when solving the mixed integer programming problem. The arranged VM table is realized by the storage device 302, for example.

  FIG. 16 is an explanatory diagram of an example of the stored contents of the arranged VM table. As illustrated in FIG. 16, the arranged VM table 1600 includes a rack number item, a server number item, a VM number item, an average CPU utilization rate item, and a standard deviation item, for each arranged VM. Configure the record.

  In the rack number item, a number for identifying the rack 203 existing in the server room is stored. The server number item stores a number for identifying the server S in the rack 203 indicated by the rack number item. In the VM number item, a number for identifying a VM that has been arranged in the server S indicated by the server S item is stored.

  The average CPU usage rate item stores the average CPU usage rate of the VM indicated by the VM number item. The average CPU usage rate may be input in advance by the user of the control server 100. In addition, the control server 100 acquires the CPU usage rate change history and execution time of the VM being executed from the server S at regular time intervals, calculates the average CPU usage rate, and updates the average CPU usage rate item. May be.

  The standard deviation item stores the standard deviation of the VM CPU utilization indicated by the VM number item. The standard deviation may be input in advance by the user of the control server 100. The control server 100 may acquire the change history and execution time of the CPU utilization rate of the VM being executed from the server S at regular time intervals, calculate the standard deviation, and update the standard deviation item.

(Functional configuration example of the control server 100)
Next, a functional configuration example of the control server 100 will be described. FIG. 17 is a block diagram illustrating a functional configuration example of the control server 100. The control server 100 includes a reception unit 1701, a specification unit 1702, a selection unit 1703, an acquisition unit 1704, a determination unit 1705, an allocation unit 1706, a detection unit 1707, and a control unit 1708. Specifically, the receiving unit 1701 to the control unit 1708 realize the functions by causing the CPU 301 to execute a program stored in the storage device 302 illustrated in FIG. 3 or the communication device 305, for example.

  The accepting unit 1701 accepts a VM allocation request to the server S transmitted from the task input server 201. Here, the allocation request includes, for example, an execution file of the VM to be allocated. Further, the allocation request may include an average CPU usage rate and standard deviation of the VMs to be allocated. When the allocation request includes the average CPU usage rate and standard deviation of the VM to be allocated, the control server 100 allocates a record that associates the VM to be allocated with the average CPU usage rate and the standard deviation. It is added to the target VM table 1500.

  The accepting unit 1701 may accept a VM placement change request currently assigned to the server S. The VM placement change request may be transmitted from the task input server 201 or may be generated in the control server 100 at regular intervals. In addition, a VM end notification from the server S may be received as a request to change the VM arrangement. Thereby, the reception unit 1701 can generate a trigger for starting the allocation of the VM to the server S. The accepted request is stored in the storage device 302, for example.

  The identifying unit 1702 identifies the second type virtual machine having the longest predicted remaining execution time from the second type virtual machine group that has been arranged in the server group. Specifically, for example, the specifying unit 1702 acquires the predicted remaining execution time of the batch VM being executed from each of the server groups, and specifies the batch VM having the longest predicted remaining execution time. In addition, the specifying unit 1702 acquires the CPU utilization rate fluctuation history and execution time of the batch VM being executed from each server group, calculates the predicted remaining execution time, and identifies the batch VM with the longest predicted remaining execution time. May be. When the batch VM includes an API that transmits information indicating the progress of the program, the specifying unit 1702 calculates the predicted remaining execution time from the information indicating the progress transmitted by the API, and the predicted remaining execution time is the longest. A batch VM may be identified.

  As a result, the specifying unit 1702 can cause the selection unit 1703 to select the server S in which the batch VM having the longest remaining prediction time has been placed as the placement destination. As a result, the specifying unit 1702 does not cause a delay in the batch VM whose predicted remaining time is not the longest, and causes the CPU resource of the server S to be freed early by the termination of the batch VM whose predicted remaining time is not the longest. Can do.

  The selection unit 1703 does not have interactive input / output with a first type virtual machine that executes predetermined processing by bidirectional communication with other devices from the server group, and performs predetermined processing by a predetermined time. An allocation destination of a virtual machine group including the second type of virtual machine that executes the standard processing to be performed is selected. Here, the first type of virtual machine is a VM that executes a predetermined process by bidirectional communication with another device, and is the above-described online VM. Further, the first type of virtual machine may be a VM to be allocated or a VM that has been arranged. The second type of virtual machine is a VM that does not have interactive input / output and executes a standard process for performing a predetermined process by a predetermined time, and is the batch VM described above. In addition, the second type of virtual machine may be a VM to be allocated or a VM that has already been arranged. A plurality of first-type virtual machines and second-type virtual machines may be selected.

  Specifically, for example, the selection unit 1703 selects, from among the server groups, a server S having a cooling efficiency that is higher by the CRAC 204 than the other servers S as an allocation destination of the VM group including the online VM and the batch VM. . The server S with good cooling efficiency by the CRAC 204 is, for example, a server S that is closer to the CRAC 204 than other servers S. As a result, the selection unit 1703 can suppress an increase in power consumption of the CRAC 204 as compared with the case where the VM is allocated to the server S having poor cooling efficiency by the CRAC 204.

  For example, the selection unit 1703 may select the server S in which the first type virtual machine has already been arranged. More specifically, for example, when there is an unallocated batch VM in the VM group, the selection unit 1703 selects the server S on which the online VM is already arranged. As a result, the online VM and the batch VM are arranged on one server S.

  In this case, the selection unit 1703 is other than the server S on which the second type virtual machine with the longest predicted remaining execution time specified by the specification unit 1702 is already placed, and the first type virtual machine has already been placed. You may select the server S which is. More specifically, the selection unit 1703 selects, for example, the server S on which the online VM has already been arranged as the server S to which the specified batch VM is assigned. As a result, the selection unit 1703 does not cause a delay in the batch VM whose predicted remaining time is not the longest due to the allocation of the online VM, and the CPU resource of the server S is quickly freed by the termination of the batch VM whose predicted remaining time is not the longest. Can occur.

  For example, the selection unit 1703 may select the server S in which the second type virtual machine has already been arranged. More specifically, for example, when there is an unallocated online VM in the VM group, the selection unit 1703 selects the server S in which the batch VM is already arranged. As a result, the selection unit 1703 assigns an online VM and a batch VM to one server S.

  In this case, the selection unit 1703 may select the server S in which the second type virtual machine having the longest predicted remaining execution time specified by the specification unit 1702 has been placed. More specifically, for example, when there is an unallocated online VM in the VM group, the selection unit 1703 selects the server S in which the specified batch VM is already arranged. As a result, the selection unit 1703 does not cause a delay in the batch VM whose predicted remaining time is not the longest due to the allocation of the online VM, and the CPU resource of the server S is quickly freed by the termination of the batch VM whose predicted remaining time is not the longest. Can occur. The selection result is stored in the storage device 302, for example.

  The acquisition unit 1704 acquires a predicted value of the arithmetic processing device usage rate based on the arithmetic processing device usage rate of the virtual machine group when the virtual machine group is assigned to the same server S. Here, the arithmetic processing device is the CPU described above. Specifically, for example, the acquisition unit 1704 acquires, from the allocation target VM table 1500, the average CPU usage rate of the batch VM that is the allocation target included in the VM group and the average CPU usage rate of the online VM that is the allocation target. To do. Then, the acquisition unit 1704 calculates the sum of the average CPU usage rate of the batch VM and the average CPU usage rate of the online VM as a predicted value.

  Here, the average CPU usage rate of the batch VM is, for example, the minimum average CPU usage rate at which the batch VM can be completed by a specified end time. The average CPU usage rate of the online VM is an average value of the CPU usage rate of the online VM. The average CPU usage rate and standard deviation of the VMs to be allocated are included in, for example, the allocation request described above, and are registered in the allocation target VM table 1500 by the control server 100. As a result, when the VM group is allocated to the server S, the acquisition unit 1704 can acquire the minimum average CPU usage rate that allows the batch VM in the VM group to be completed by the specified end time.

  Specifically, the acquisition unit 1704 acquires, for example, the average CPU usage rate of the batch VMs to be allocated included in the VM group from the allocation target VM table 1500. Next, the acquisition unit 1704 acquires the average CPU usage rate of the arranged online VMs included in the VM group from the arranged VM table 1600. Then, the acquisition unit 1704 calculates the sum of the average CPU usage rate of the batch VM and the average CPU usage rate of the online VM as a predicted value.

  Here, the average CPU usage rate and standard deviation of the allocated VM are registered in the allocated VM table 1600 when, for example, the control server 100 has previously allocated the allocation target VM to any server S. ing. In this case, the record related to the allocated VM in the allocation target VM table 1500 is deleted. As a result, when the acquisition unit 1704 allocates the batch VM to be allocated to the server S on which the online VM has been allocated, the minimum average CPU utilization that can end the batch VM to be allocated by the specified end time. Rate can be obtained.

  Specifically, the acquisition unit 1704 acquires, for example, the average CPU usage rate of the arranged batch VMs included in the VM group from the arranged VM table 1600. Next, the acquisition unit 1704 acquires the average CPU usage rate of the online VM that is the allocation target included in the VM group from the allocation target VM table 1500. Then, the acquisition unit 1704 calculates the sum of the average CPU usage rate of the batch VM and the average CPU usage rate of the online VM as a predicted value. As a result, when the acquisition unit 1704 allocates an online VM to be allocated to the server S in which the batch VM has been allocated, the minimum average CPU usage rate that can end the allocated batch VM by the specified end time. Can be obtained.

  Specifically, the acquisition unit 1704 acquires, for example, the average CPU usage rate of the online VM and σ × n of the CPU usage rate of the online VM from the allocation target VM table 1500. Then, the acquiring unit 1704 calculates the sum of the average CPU usage rate of the online VM and the CPU usage rate σ × c of the online VM as a predicted value. Here, σ is a standard deviation in the normal distribution of the CPU usage rate of the online VM. n is a constant for multiplying the standard deviation. As a result, the acquisition unit 1704 can acquire the CPU usage rate when the online VM to be allocated becomes a high load.

  Specifically, the acquisition unit 1704 acquires, for example, the average CPU usage rate of the online VM and σ × n of the CPU usage rate of the online VM from the arranged VM table 1600. Then, the acquiring unit 1704 calculates the sum of the average CPU usage rate of the online VM and the CPU usage rate σ × c of the online VM as a predicted value. As a result, the acquisition unit 1704 can acquire the CPU usage rate when the placed online VM becomes highly loaded.

  The sum of the calculated average CPU usage rate of the plurality of batch VMs of the VM group and the average CPU usage rate of the plurality of online VMs of the VM group is stored in the storage device 302, for example. Also, the sum of the average CPU usage rate of the online VM and the CPU usage rate σ × n of the online VM is also stored in the storage device 302, for example.

The determination unit 1705 determines whether or not the predicted value acquired by the acquisition unit 1704 is equal to or less than a threshold value based on the number of arithmetic processing devices of the server S selected by the selection unit 1703. The threshold value is the upper limit of the CPU usage rate in the server S, and is the aforementioned _Pi . P _i is a value obtained by multiplying the upper limit of the CPU utilization of the number and 1CPU per CPU in the server S. The P _i, for example, (number × 90% of the CPU) - 10% is employed.

Specifically, for example, the determination unit 1705 determines whether the sum of the average CPU usage rate of the plurality of batch VMs in the VM group and the average CPU usage rate of the plurality of online VMs in the VM group is _Pi or less. Judge whether or not. As a result, the determination unit 1705 determines the minimum average CPU usage rate that allows each of the batch VMs to be completed by the specified end time even if the CPU usage rates of the multiple online VMs are considered. It can be determined whether or not the CPU utilization rate upper limit can be secured.

Further, determination unit 1705, specifically, for example, an average CPU utilization line VM, and sigma × n of the CPU utilization of online VM, the sum of, determining whether a P _i less . Accordingly, the determination unit 1705 determines whether or not the sum of the CPU utilization rates of the online VMs is equal to or less than the upper limit of the CPU utilization rate of the server S, even when a plurality of online VMs are simultaneously loaded. can do. The determination result is stored in the storage device 302, for example.

  When the determination unit 1705 determines that the allocation unit 1706 is equal to or less than the threshold value, the allocation unit 1706 assigns the virtual machine group to the selected server S so that the first type virtual machine is preferentially executed over the second type virtual machine. assign. Specifically, for example, when the determination unit 1705 determines that the allocation unit 1706 is equal to or less than the threshold value, the allocation unit 1706 transmits a VM group execution request to the selected server S. Here, the execution request includes each execution file of the VM group. The execution request includes a notification indicating that the priority of execution of the online VM in the VM group is higher than that of the batch VM. As a result, when the allocation unit 1706 allocates the batch VM and the online VM to be allocated to the server S, the allocation unit 1706 avoids the performance degradation of the allocated online VM and ends the allocated batch VM by the specified end time. be able to.

  The execution request may include an instruction for causing each VM group to be executed by the same CPU of the server S. As a result, the allocation unit 1706 transmits an execution request including an instruction to execute each VM group by the same CPU of the server S to the server S, thereby causing the CPU group to execute the VM group in the server S. CPU resources can be used efficiently.

  When the allocating unit 1706 allocates the VM group to the server S, the allocating unit 1706 refers to the record regarding each VM group in the allocation target VM table 1500 and registers the record regarding each VM group in the arranged VM table 1600. In addition, the allocating unit 1706 deletes the record relating to each VM group in the allocation target VM table 1500 after registering in the arranged VM table 1600. Also, the assignment unit 1706 updates a record related to the assigned server S in the initial arrangement table 1300. Thereby, when allocating a new VM, the allocation unit 1706 can acquire the average CPU usage rate and standard deviation of the VM allocated by the allocation unit 1706 from the allocated VM table 1500.

  In addition, for example, when the server S in which the batch VM is already arranged is selected, the assignment unit 1706 transmits an online VM execution request to the selected server S, for example. Here, the execution request includes a notification for making the execution priority of the online VM subject to the execution request higher than the batch VM already arranged in the selected server S. In accordance with the notification, the server S uses CPU resources preferentially for execution of online VMs, and uses excess CPU resources not used for execution of online VMs for execution of batch VMs. As a result, when the allocation unit 1706 allocates the online VM to be allocated to the server S to which the batch VM has been allocated, the allocation unit 1706 avoids the performance degradation of the allocated online VM, and sets the allocated batch VM to the specified end time. Can be finished by

  Further, the execution request may include an instruction to execute the allocated online VM by the CPU in which the batch VM of the server S is already arranged. As a result, the allocating unit 1706 can cause the server S to execute the online VM to be allocated by the CPU on which the batch VM has already been allocated, and allow the server S to efficiently use CPU resources.

  In addition, for example, when the server S in which the online VM has already been selected is selected, the allocating unit 1706 transmits a batch VM execution request to the selected server S, for example. In accordance with the notification, the server S uses CPU resources preferentially for execution of online VMs, and uses excess CPU resources not used for execution of online VMs for execution of batch VMs. As a result, when the allocation unit 1706 allocates the batch VM to be allocated to the server S to which the online VM has been allocated, the allocation unit 1706 avoids the performance degradation of the allocated online VM and sets the allocated batch VM to the specified end time. Can be finished by

  Further, the execution request may include an instruction to execute the assigned batch VM by the CPU on which the online VM of the server S has been arranged. As a result, the allocating unit 1706 can cause the CPU on which the online VM of the server S has been allocated to execute the batch VM to be allocated and allow the server S to efficiently use CPU resources.

  The detection unit 1707 detects a server S to which no VM is assigned from the server group. Specifically, for example, the detection unit 1707 requests VM execution statuses from each of the server S groups at regular time intervals, and acquires any VM execution status by acquiring the VM execution status of each server S group. An unassigned server S is detected. In addition, the detection unit 1707 may acquire the execution status of the VM of the server S in which the allocated VM is reallocated to another server S as a result of the allocation by the allocation unit 1706. The detection result is stored in the storage device 302, for example.

  The control unit 1708 performs control so that the power consumption of the server S to which no virtual machine is assigned is reduced. Specifically, for example, the control unit 1708 transmits a request for stopping power supply to the server S detected by the detection unit 1707 to the UPS 206 that controls power supply to the server S. Accordingly, the control unit 1708 can reduce the power consumption of the server S.

  The functions of the selection unit 1703, the acquisition unit 1704, and the determination unit 1705 can be realized by solving a mixed integer programming problem that includes the above formulas (1) and (2) as constraints. A solver may also be used when solving a mixed integer programming problem. Examples of solvers that solve the mixed integer programming problem include GLPK, SYMPHONY, Gurobi, and CPLEX.

  Specifically, for example, the mixed integer programming problem is expressed by the minimizing function of the following equation (5) and the constraint conditions of the following equations (6) to (15). The values of the variables in the mixed integer programming problems (5) to (15) are stored in the storage device 502 in advance, for example.

  Equation (5) above shows the minimizing function in the mixed integer programming problem. The control server 100 can obtain a solution that minimizes the minimizing function within the range satisfying the constraint conditions of the following formulas (6) to (15) by solving the mixed integer programming problem. The above formula (5) specifically indicates the power consumption in the server room. Therefore, the control server 100 can obtain a solution that minimizes the power consumption in the server room by solving the mixed integer programming problem.

  Here, N is the number of racks 203 in the server room. i is the number of the rack 203 and is a number from 1 to N. S is the number of servers S in the rack 203. j is the number of the server S of the rack 203, and is a number from 1 to S. M is the number of VMs to be allocated. n is a VM number, which is a number from 1 to M. C is the number of CRACs 204. k is the number of the CRAC 204 and is a number from 1 to C.

η _i is the power that is constantly consumed by the UPS 206 of the rack 203 (i). ε _i is a proportionality coefficient indicating how many times the UPS 206 of the rack 203 (i) consumes the power of the server S of the rack 203 (i). α _ij is the power that is constantly consumed by the server S (i, j). v _ij is a variable indicating whether the power source of the server S (i, j) is ON or OFF, and is “1” when the server S (i, j) is ON, and “0” when it is OFF.

β _ij is the power consumption per VM in the server S (i, j). W _ij ⁰ is the number of VMs already arranged in the server S. τ _n is the number of VMs to be allocated. x _{n, i, j} is a variable indicating whether or not VM (n) is assigned to the server S (i, j), and becomes “1” when VM (n) is assigned. It is “0” when n) is not assigned.

λ _k is the power that is constantly consumed by the CRAC 204 (k). μ _k is a proportionality coefficient indicating how many times power is consumed by the CRAC 204 (k) with respect to the heat flow rate flowing into the CRAC 204 (k). q _{in, CRAC} ^K is a heat flow rate flowing into the _CRAC 204 (k). kappa _ij is, Chiller205 (k) is, with respect to the heat flow flowing into Chiller205 (k), is a proportionality coefficient that indicates to consume many times the power. q _{out, CRAC} ^K is a heat flow rate flowing _{out of CRAC} 204 (k).

More specifically, Σ _{j = 1} ^S α _ij v _ij in the above equation (5) represents the sum of the constant power consumption of the plurality of servers S in the rack 203 (i). In the above equation (5), Σ _{j = 1} ^S β _ij W _ij ⁰ represents the sum of the power consumption of the VMs already arranged in the plurality of servers S in the rack 203 (i). Σ _{j = 1} ^S (β _ij Σ _{n = 1} ^M τ _n x _{n, i, j} ) in the above equation (5) is the power consumption of the VM allocated to the plurality of servers S in the rack 203 (i). Indicates the sum. Therefore, Σ _{j = 1} ^S (α _ij v _ij + β _ij (W _ij ⁰ + Σ _{n = 1} ^M τ _n x _{n, i, j} )) in the above equation (5) is _{equal to} a plurality of racks 203 (i). The total power consumption of the server S is shown.

In the above equation (5), ε _i (Σ _{j = 1} ^S (α _ij v _ij + β _ij (W _ij ⁰ + Σ _{n = 1} ^M τ _n x _{n, i, j} ))) in the rack 203 (i) The power consumption of the UPS 206 is shown. Therefore, Σ _{i = 1} ^N (η _i + (1 + ε _i ) (Σ _{j = 1} ^S (α _ij v _ij + β _ij (W _ij ⁰ + Σ _{n = 1} ^M τ _n x _{n, i, j} )))) The sum of power consumption of a plurality of racks 203 in the server room is shown.

In addition, Σ _{k = 1} ^C (λ _k + μ _k q _{in, CRAC} ^K ) _in the above equation (5) represents the sum of power consumed by _{the CRAC} 204 for cooling the server S. Σ _{k = 1} ^C (κ _k q _{out, CRAC} ^K ) in the above equation (5) indicates the sum of electric power consumed by the chiller 205 for exhausting the heat flow that has flowed _{out of the CRAC} 204. Accordingly, Σ _{k = 1} ^C (λ _k + μ _k q _{in, CRAC} ^K ) + Σ _{k = 1} ^C (κ _k q _{out, CRAC} ^K ) indicates the total power consumption for cooling the server S in the server room. .

  Therefore, the above formula (5) indicates the total power consumption of the server room. Therefore, the control server 100 can obtain a solution that minimizes power consumption in the server room by solving the integer programming problem.

In the above equation (6), the variable x _{n, i, j} indicating whether or not VM _n is assigned to the server S (i, j) is a discrete value that takes a binary value of “0” or “1”. It is a constraint condition indicating that it is a variable. The variable x _{n, i, j} indicating whether or not VM _n is assigned to the server S (i, j) does not take a value other than “0,1” due to the constraint condition of the above equation (6). To.

The above equation (7) indicates that the variable v _ij indicating whether the power source of the server S (i, j) is ON or OFF is a discrete variable that takes a binary value of “0” or “1”. It is a constraint condition. The variable v _ij indicating whether the power source of the server S (i, j) is in an ON state or an OFF state is prevented from taking a value other than “0,1” by the constraint condition of the above equation (7).

The above equation (8) is a constraint condition indicating that the variable θ _i indicating whether or not the VM is allocated to the rack 203 (i) is a discrete variable that takes a binary value of “0” or “1”. It is. The variable θ _i indicating whether or not the VM is allocated to the rack 203 (i) is prevented from taking a value other than “0, 1” by the constraint condition of the above equation (8).

The above equation (9) is a constraint indicating that the sum of the variables x _{n, i, j} related to each server S in the server S group in the server room is “1”. Therefore, among each server S, the variable x _{n, i, j} related to any one server S becomes “1”, and the variable x _{n, i, j} related to the remaining server S becomes “0”. In other words, the constraint condition of the above formula (9) indicates that VM _n is assigned to any one server S, and there is no case where it is assigned to two or more servers S or not assigned to servers S. It shows that.

The above equation (10) is a constraint condition corresponding to the above equation (1), and indicates a condition for ending the batch VM by the specified end time. Here, M _{0, i, j} is the number of allocated VMs. τ _{0, i, j, m} is the average CPU utilization of the VM (m) that has already been placed in the server S (i, j). τ _n is the average CPU usage rate of the VM (n) to be allocated. γ _ij is a proportional coefficient indicating how many times the performance of the server S (i, j) is higher than that of the reference server S. L _ij is the upper limit of the CPU usage rate of the server S (i, j).

  Specifically, the condition of the above formula (10) is the minimum that can finish each of the plurality of batch VMs by the predetermined end time even in consideration of the average CPU utilization rate of each of the plurality of online VMs. It is shown that the average CPU usage rate can be secured below the CPU usage rate upper limit of the server S. The batch VM can be completed by the specified end time due to the constraint condition of the above equation (1).

The above equation (11) is a constraint equation corresponding to the above equation (2), and indicates a condition for avoiding the performance degradation of the online VM. Here, a _{0, i, j, m} is the number of online VMs already placed in the server S (i, j), and is a number less than or equal to M _{0, i, j} . a is the number of online VMs to be allocated, and is a number equal to or less than M.

σ _{0, i, j, m} is a standard deviation in the normal distribution of the CPU utilization rate of the online VM (m) that has already been arranged in the server S (i, j). c is a constant for multiplying the standard deviation. Therefore, τ _{0, i, j, m} + σ _{0, i, j, m} × c is the CPU utilization rate when the online VM (m) placed in the server S (i, j) is simultaneously subjected to a high load. It is.

σ _n is a standard deviation in the normal distribution of the CPU usage rate of the online VM (n) to be allocated. Therefore, x _{n, i, j} (τ _n + σ _n × c) is the CPU utilization rate when the online VM (n) to be allocated becomes a high load.

  Specifically, the condition of the above formula (2) is that the sum of the CPU utilization rates of the online VMs is equal to or less than the upper limit of the CPU utilization rate of the server S even when a plurality of online VMs are simultaneously subjected to a high load. It shows that it becomes. Due to the constraint condition of the above equation (2), competition between online VMs is avoided and performance degradation of online VMs is avoided.

  The above equation (12) is a constraint condition indicating that the number of VMs allocated to the rack 203 (i) is equal to or less than the number M of VMs to be allocated.

  The above formulas (13) to (15) are constraints regarding the cooling state in the server room. Due to the constraints of the above formulas (13) to (15), it is possible to eliminate the server S with insufficient cooling. The values of the variables in the above formulas (13) to (15) are stored in the storage device 502, for example.

Further, the control server 100 can reduce the total power consumption of the server S in the server room without considering the power consumption of the CRAC 204 in the server room. In this case, for the mixed integer programming problem, Σ _{k = 1} ^C (λ _k + μ _k q _{in, CRAC} ^K ) + Σ _{k = 1} ^C (κ _k q _{out, CRAC} ^K ) in the above equation (5) and the above equation ( 13) to (15) are deleted.

(Functional configuration example of server S)
Next, a functional configuration example of the server S will be described. FIG. 18 is a block diagram illustrating a functional configuration example of the server S. The server S is configured to include a reception unit 1801, a determination unit 1802, an allocation unit 1803, and a control unit 1804. The accepting unit 1801 to the control unit 1804 realize the functions by causing the CPU 301 to execute the program stored in the storage device 302 illustrated in FIG. 3 or the communication device 305.

  Specifically, the reception unit 1801 receives, for example, a VM group execution request transmitted from the control server 100 to the own device. Here, the execution request includes each execution file of the VM group. The execution request includes a notification indicating that the priority of execution of the online VM in the VM group is higher than that of the batch VM.

  More specifically, the accepting unit 1801 accepts, for example, execution requests for an online VM and a batch VM that are assigned to the own device and are executed simultaneously. Here, the online VM and the batch VM are VM groups that can satisfy the constraint conditions of the above formulas (10) and (11) when executed simultaneously in the own apparatus. Further, the execution request may include information for designating a CPU to which the VM group is arranged. Thereby, the reception unit 1801 triggers to start the processing of the online VM that can avoid the performance degradation and the batch VM that can be completed by the specified end time even if the reception unit 1801 executes simultaneously in the own device. Can be generated.

  The accepting unit 1801 may accept an execution request for a batch VM that is executed at the same time as an online VM that has already been placed in the own device. In addition, the execution request may include information for designating a CPU in which an online VM that is the destination of the VM group has already been placed. As a result, even if the reception unit 1801 executes the online VM that has already been arranged in its own device, the online VM that has already been arranged does not cause performance degradation, and the batch VM that can be terminated by the specified end time. A trigger to start processing can be generated.

  In addition, the reception unit 1801 may receive an execution request for an online VM that is executed simultaneously with the batch VM that has been arranged in the own device. In addition, the execution request may include information for designating a CPU in which a batch VM that is a placement destination of the VM group has been placed. Here, the online VM and the batch VM are VM groups that can satisfy the constraint conditions of the above formulas (10) and (11) when executed simultaneously in the own apparatus. As a result, even if the reception unit 1801 executes the batch VM at the same time as the allocated batch VM in its own device, the allocated batch VM is completed by the specified end time, and the online VM that can avoid performance degradation can be avoided. A trigger to start processing can be generated.

  The determining unit 1802 determines the placement destination of the virtual machine group as one of the plurality of arithmetic processing devices so as to be within the processing capability of the arithmetic processing device. Here, the virtual machine group includes a first type virtual machine that executes a predetermined process by bidirectional communication with another device, and a fixed process that performs a predetermined process by a predetermined time without interactive input / output And a second type of virtual machine. The arithmetic processing unit is a CPU. Specifically, the determination unit 1802 determines, for example, the placement destination of the VM group as a CPU in which a VM has been placed among a plurality of CPUs. In addition, when the execution request includes information specifying the placement destination CPU, the determination unit 1802 may determine the CPU indicated by the information as the placement destination. Thereby, the determination unit 1802 can execute a VM group by the same CPU and efficiently use CPU resources.

  In addition, when the first type virtual machine has already been arranged in any one of the plurality of arithmetic processing devices, the determination unit 1802 determines the placement type of the second type virtual machine as the first type virtual machine. You may determine to the arithmetic processing apparatus by which the virtual machine has already been arrange | positioned. Specifically, the determination unit 1802 determines, for example, the placement destination of the batch VM as the CPU on which the online VM has been placed. As a result, the determination unit 1802 executes, with the same CPU, the arranged online VM that can avoid performance degradation even when executed simultaneously and the batch VM to be allocated that can be completed by the specified end time. Can be made.

  In addition, when the second type virtual machine has already been placed in any one of the plurality of processing units, the determination unit 1802 determines the placement type of the first type virtual machine as the second type. You may determine to the arithmetic processing apparatus by which the virtual machine has already been arrange | positioned. Specifically, for example, the determination unit 1802 determines the placement destination of the online VM as the CPU on which the batch VM has been placed. As a result, the determination unit 1802 executes, with the same CPU, an online VM that is an allocation target that can avoid performance degradation even if it is executed at the same time, and an already placed batch VM that can be ended by the specified end time. Can be made.

  The assigning unit 1803 assigns a virtual machine group to the arithmetic processing device determined as the placement destination by the determining unit 1802 among the plurality of arithmetic processing devices. Thereby, the allocation unit 1803 can cause the CPU to start the VM processing.

  The control unit 1804 performs control so that the first type virtual machine is executed with priority over the second type virtual machine. Specifically, for example, the control unit 1804 notifies the CPU so that the priority of the online VM is higher than the priority of the batch VM. For this reason, the CPU uses CPU resources preferentially for execution of online VMs, and uses extra CPU resources not used for execution of online VMs for execution of batch VMs. As a result, the control unit 1804 can avoid the performance degradation of the online VM.

(Assignment result of VM to server S by control server 100)
Next, VM allocation results to the server S by the control server 100 will be described with reference to FIGS. 19 and 20.

  FIGS. 19 and 20 are explanatory diagrams illustrating an example of a VM allocation result to the server S by the control server 100. FIG. In the example of FIG. 19, the online VM 11, the batch VM 12, the online VM 13, the batch VM 14, the online VM 15, the online VM 16, and the batch VM 17 are already arranged in the server S 1. In addition, the batch VM 18 has already been placed in the server S2. The server S3 is not assigned a VM and is not supplied with power. As an example, the average CPU usage rate of the online VM is 30%, the CPU usage rate including the standard deviation of the online VM is 80%, the average CPU usage rate of the batch VM is 30%, and the threshold value of each server S is 360%.

  The control server 100 assigns the online VM 19 to any server S. The control server 100 acquires the average CPU usage rate and the standard deviation of the CPU usage rate of the online VM 19 from the allocation target VM table 1500, and the average CPU usage rate and the CPU usage rate of the allocated VMs 11 to 18 from the allocated VM table 1600. Get the standard deviation of. Then, the control server 100 solves the mixed integer programming problem using the acquired average CPU utilization rate and standard deviation.

Specifically, the control server 100 inputs, for example, a mixed integer programming problem in which values of variables stored in the respective tables illustrated in FIGS. 4 to 16 are set to the GLPK, and the mixed integer programming problem obtained by the GLPK. Get the solution of For example, the GLPK outputs the variable x _{n, i, j} when the VM 19 is allocated as a solution.

  For example, in any of the servers S1 to S3, the threshold that is the sum of the average CPU usage rate of the arranged VMs and the average CPU usage rate “30%” of the online VM 19 is the upper limit of the CPU usage rate “360%” or less. Therefore, the servers S1 to S3 satisfy the condition of the above formula (10).

  Further, in the server S1, the sum of the CPU utilization rate including the standard deviation of the arranged online VM “320%” and the CPU utilization rate including the standard deviation of the online VM 19 “80%” is the CPU utilization rate. The upper threshold “360%” is exceeded. Therefore, the server S1 does not satisfy the condition of the above formula (11). In addition, in the servers S2 and S3, the sum of the CPU usage rate including the standard deviation of the online VMs already arranged and the CPU usage rate “80%” including the standard deviation of the online VM 19 is the upper limit of the CPU usage rate. The threshold is “360%” or less. Therefore, the servers S2 and S3 satisfy the condition of the above formula (11).

Further, when the online VM 19 is allocated to the server S3 to which power is not supplied, power is supplied to the server S3, and the online VM 19 is allocated and executed. In this case, power consumption in the server room increases, and the minimizing function of the above equation (5) cannot be minimized. Therefore, for example, the GLPK sets the variable x _19,1,2 indicating whether or not the online VM 19 is allocated to the server S2 of the rack 203 (1) to “1” and outputs it.

Then, the control server 100 allocates the online VM 19 to the server S2 corresponding to the output solution “x _19,1,2 = 1”. Thereby, the control server 100 can avoid the performance degradation of online VM. In addition, the control server 100 can allocate VMs so that the batch VMs are completed by the specified end time and the number of servers S to which VMs are allocated decreases.

  Further, when the online VM 19 is allocated, the control server 100 refers to the record related to the online VM 19 in the allocation target VM table 1500 and adds the record related to the online VM 19 to the arranged VM table 1600. Also, the control server 100 deletes the record related to the online VM 19 in the allocation target VM table 1500. In addition, the control server 100 updates the initial arrangement item regarding the server S2 in the initial arrangement table 1300. Thereby, the control server 100 can acquire the average CPU usage rate of the online VM 19 and the standard deviation of the CPU usage rate from the arranged VM table 1600 when a new VM is allocated thereafter.

  On the other hand, the server S2 can avoid the performance degradation of the online VM in the already-arranged VM group and can finish the batch VM in the already-arranged VM group by the specified end time. Then, the online VM 19 is executed by any CPU. Thereby, the server S2 can avoid the performance degradation of the online VM and can finish the batch VM by the specified end time.

  Next, in the example of FIG. 20, the online VM 19 assigned to the server S2 has ended. Here, when it is detected that the VM assigned to the server S2 is the batch VM 18, the control server 100 tries whether the batch VM 18 can be migrated to another server S.

  Specifically, the control server 100 excludes the batch VM 18 from the arranged VMs, solves the mixed integer programming problem when the batch VM 18 is adopted as the allocation target VM, and reassigns the batch VM 18.

  Specifically, the control server 100 refers to a record related to the batch VM 18 in the arranged VM table 1600 and adds a record related to the batch VM 18 to the allocation target VM table 1500. Further, the control server 100 deletes the record related to the batch VM 18 in the arranged VM table 1600. In addition, the control server 100 updates the initial arrangement items regarding the servers S1 and S2 in the initial arrangement table 1300. Thereby, the control server 100 can handle the batch VM 18 as a VM to be assigned.

Then, the control server 100 inputs the mixed integer programming problem in which the values of the variables stored in the tables shown in FIGS. 4 to 16 are set to the GLPK, and acquires the solution of the mixed integer programming problem obtained by the GLPK. . For example, GLPK outputs the variable x _{n, i, j} when the VM 18 is allocated as a solution. Then, the control server 100 assigns the batch VM 18 to the server S corresponding to the solution of the mixed integer programming problem.

  For example, in any of the servers S1 to S3, a threshold that is the sum of the average CPU usage rate of the arranged VMs and the average CPU usage rate “30%” of the batch VM 18 is the upper limit of the CPU usage rate. “360%” or less. Therefore, the servers S1 to S3 satisfy the condition of the above formula (10).

  In any of the servers S1 to S3, the CPU usage rate including the standard deviation of the arranged online VM is equal to or less than the threshold “360%” that is the upper limit of the CPU usage rate. Therefore, the servers S1 to S3 satisfy the condition of the above formula (11).

Further, when the batch VM 18 is not assigned to the server S2, no VM is assigned to the server S2. In this case, if not assigned to the server S2, the control server 100 can stop the power supply to the server S2. Thereby, power consumption in the server room is reduced, and the minimizing function of the above equation (5) is reduced. When the batch VM 18 is assigned to the server S3, the control server 100 supplies power to the server S3 to execute the batch VM 18. In this case, power consumption in the server room increases, and the minimizing function of the above equation (5) cannot be minimized. From the above, the GLPK outputs, for example, the variable x _19,1,1 indicating whether the batch VM 18 is allocated to the server S1 of the rack 203 (1) as “1”.

Here, the control server 100 assigns the batch VM 18 to the server S1 corresponding to the solution “x _19,1,1 = 1” of the mixed integer programming problem, and migrates the batch VM 18 from the server S2. When the control server 100 detects that there is no VM assigned to the server S2, the control server 100 requests the UPS 206 to stop supplying power to the server S2. Thereby, the power consumption of a server room is reduced. Further, the control server 100 also migrates the VM to the server S that satisfies the conditions of the above formulas (10) and (11) in the migration of the VM, so the performance degradation of the online VM is avoided and the batch VM is specified. It can be finished by time.

  In addition, if the server S corresponding to the solution of the mixed integer programming problem is the server S2, the control server 100 does not perform VM migration. Thereby, if there is a migration destination that can reduce power consumption, the control server 100 can reduce the power consumption of the server room by migrating the VM. In general, since the VM performance deteriorates during the migration, the control server 100 delays the response to the client 202 when the online VM is migrated, and the user of the client 202 suffers a disadvantage or an SLA violation. May occur. Therefore, the control server 100 may not perform the migration for the online VM.

(Assignment process)
Next, a detailed processing procedure of the allocation processing executed by the control server 100 will be described using FIG.

  FIG. 21 is a flowchart showing a detailed processing procedure of the allocation processing. In FIG. 21, first, the control server 100 determines whether or not a VM allocation request has been received (step S2101). If a VM allocation request has not been received (step S2101: No), the control server 100 returns to step S2101 and waits for an allocation request.

  On the other hand, when a VM allocation request is received (step S2101: Yes), the control server 100 determines whether the allocation-requested VM is an online VM (step S2102). Here, in the case of an online VM (step S2102: Yes), the control server 100 acquires the average CPU usage rate of the online VM and the standard deviation of the CPU usage rate of the online VM from the allocation target VM table 1500. (Step S2103). Then, the control server 100 proceeds to Step S2105.

  On the other hand, when it is not an online VM (step S2102: No), the control server 100 acquires the average CPU usage rate of the batch VM from the allocation target VM table 1500 (step S2104). Then, the control server 100 proceeds to Step S2105.

  Next, the control server 100 solves the mixed integer programming problem using the acquired average CPU usage rate of the batch VM or the acquired average CPU usage rate and standard deviation of the online VM (step S2105).

  Next, the control server 100 allocates VMs according to the solution of the mixed integer programming problem (step S2106). Then, the control server 100 ends the allocation process. Thereby, the control server 100 can allocate the VM so that the performance degradation of the online VM is avoided, the batch VM is completed by the specified end time, and the number of servers S to which the VM is allocated is reduced. .

  As described above, the allocating device is based on the condition that the sum of the average CPU usage rate of the online VM and the CPU usage rate σ × c of the online VM does not exceed the upper limit of the CPU usage rate in the server S. Then, assign a VM. Thereby, the allocation apparatus can avoid the competition between online VMs in the server S, and can avoid the performance degradation of online VMs.

  In the allocation device, the sum of the average CPU utilization rate of the plurality of batch VMs in the VM group and the average CPU utilization rate of the plurality of online VMs in the VM group does not exceed the upper limit of the CPU utilization rate in the server S. Assign VMs under conditions. Therefore, the allocating device can allocate the VM so that the batch VM ends by the specified end time.

  Further, the allocating device can reduce the number of servers S to which VMs are allocated in the server room by collectively allocating the VM group to one server S. As a result, if there is a server S to which no VM is allocated, the allocation device can stop power supply and reduce power consumption of the server room.

  In addition, the allocating device can reduce the number of servers S to which the VMs are allocated in the server room by transferring the allocated VMs. As a result, if there is a server S to which no VM is allocated, the allocation device can stop power supply and reduce power consumption of the server room.

  In this case, when the allocated VM is an online VM, the allocating apparatus can avoid the performance degradation of the online VM by not using the online VM as a migration target.

  The allocation method and information processing method described in this embodiment can be realized by executing a program prepared in advance by a computer such as a personal computer or a workstation. The allocation program and the information processing program are recorded on a computer-readable recording medium such as a hard disk, a flexible disk, a CD-ROM, an MO, and a DVD, and are executed by being read from the recording medium by the computer. Further, the allocation program and the information processing program may be distributed through a network such as the Internet.

  The following additional notes are disclosed with respect to the embodiment described above.

(Supplementary Note 1) A first type of virtual machine that executes a predetermined process by bidirectional communication with another device, and a standard process that performs a predetermined process by a predetermined time without interactive input / output A reception unit that receives the virtual machine group from which the predicted value of the arithmetic processing unit usage rate of the virtual machine group including the two types of virtual machines is equal to or less than a threshold value;
A determination unit that determines an arrangement destination of the virtual machine group received by the reception unit to be any one of the plurality of arithmetic processing devices so as to be within the processing capacity of the arithmetic processing device;
An allocating unit for allocating the virtual machine group to the arithmetic processing device determined as the placement destination by the determining unit among the plurality of arithmetic processing devices;
A control unit that controls the first type virtual machine to be executed in preference to the second type virtual machine;
An information processing apparatus comprising:

(Supplementary Note 2)
When the first type virtual machine has already been placed in any one of the plurality of processing units, the second type virtual machine has been placed in the first type virtual machine. The information processing apparatus according to appendix 1, wherein the information processing apparatus is determined to be an arithmetic processing apparatus.

(Supplementary note 3)
When the second type virtual machine has already been placed in any one of the plurality of processing units, the second type virtual machine has been placed as the placement destination of the first type virtual machine. The information processing apparatus according to appendix 1, wherein the information processing apparatus is determined to be an arithmetic processing apparatus.

(Supplementary Note 4) A first type of virtual machine that executes predetermined processing by bidirectional communication with another device from the server group, and does not have interactive input / output, and performs predetermined processing by a predetermined time A selection unit that selects an assignment destination of a virtual machine group including a second type of virtual machine that executes the standard processing;
An acquisition unit that acquires a predicted value of the arithmetic processing device usage rate based on the arithmetic processing device usage rate of the virtual machine group when the virtual machine group is assigned to the same server;
A determination unit that determines whether or not the predicted value acquired by the acquisition unit is equal to or less than a threshold value based on the number of arithmetic processing devices of the server selected by the selection unit;
When the determination unit determines that the virtual machine group is less than or equal to the threshold value, the virtual machine group is set to the selected server so that the first type virtual machine is executed in preference to the second type virtual machine. Assigning part to be assigned,
An allocating device comprising:

(Supplementary note 5)
Select the server on which the first type virtual machine has been placed,
The allocation unit is
When the determination unit determines that the threshold value is less than or equal to the threshold value, the second type is sent to the selected server so that the first type virtual machine is executed in preference to the second type virtual machine. The assigning device according to appendix 4, wherein the virtual machine is assigned.

(Additional remark 6) It has the specific part which specifies the said 2nd type virtual machine with the longest prediction remaining execution time from the said 2nd type virtual machine group already arrange | positioned at the said server group,
The selection unit includes:
Selecting a server other than the server on which the second type virtual machine with the longest predicted remaining execution time specified by the specifying unit is placed, and on which the first type virtual machine has been placed. Item 6. The allocation device according to appendix 5.

(Supplementary note 7)
Select the server on which the second type of virtual machine has been placed,
The allocation unit is
The supplementary note 4 is characterized in that the first type virtual machine is allocated to the selected server so that the first type virtual machine is executed in preference to the second type virtual machine. Allocation device.

(Additional remark 8) It has the specific part which specifies the said 2nd type virtual machine with the longest prediction remaining execution time from the said 2nd type virtual machine group already arrange | positioned at the said server group,
The selection unit includes:
8. The allocation apparatus according to appendix 7, wherein a server in which the second type virtual machine having the longest predicted remaining execution time specified by the specifying unit is already arranged is selected.

(Additional remark 9) The control part which controls so that the power consumption of the server to which any virtual machine is not allocated among the said server groups falls is provided. Any one of additional remark 4-8 characterized by the above-mentioned. Allocation device.

(Supplementary note 10)
A first type of virtual machine that executes a predetermined process by bidirectional communication with another device, and a second type of virtual machine that does not have interactive input / output and executes a standard process that performs the predetermined process by a predetermined time A virtual machine group including the machine, and accepting the virtual machine group from which the predicted value of the arithmetic processing unit usage rate of the virtual machine group is equal to or less than a threshold value from the allocation device,
The placement destination of the accepted virtual machine group is determined to be any one of the plurality of arithmetic processing devices so as to be within the processing capacity of the arithmetic processing device,
Assign the virtual machine group to the arithmetic processing device determined as the placement destination among the plurality of arithmetic processing devices,
Controlling the first type of virtual machine to be preferentially executed over the second type of virtual machine;
An information processing method characterized by executing processing.

(Appendix 11) The computer
From the server group, the first type of virtual machine that executes predetermined processing by bidirectional communication with other devices and routine processing that performs predetermined processing by a predetermined time without interactive input / output Select the assignment destination of the virtual machine group including the second type of virtual machine
Obtaining a predicted value of the processing unit usage rate based on the processing unit usage rate of the virtual machine group when the virtual machine group is assigned to the same server;
Determining whether the obtained predicted value is less than or equal to a threshold value based on the number of processing units of the selected server;
If the virtual machine group is determined to be equal to or lower than the threshold value, the virtual machine group is assigned to the selected server so that the virtual machine of the first type is executed in preference to the virtual machine of the second type.
An allocation method characterized by executing processing.

(Supplementary note 12)
A first type of virtual machine that executes a predetermined process by bidirectional communication with another device, and a second type of virtual machine that does not have interactive input / output and executes a standard process that performs the predetermined process by a predetermined time A virtual machine group including the machine, and accepting the virtual machine group from which the predicted value of the arithmetic processing unit usage rate of the virtual machine group is equal to or less than a threshold value from the allocation device,
The placement destination of the accepted virtual machine group is determined to be any one of the plurality of arithmetic processing devices so as to be within the processing capacity of the arithmetic processing device,
Assign the virtual machine group to the arithmetic processing device determined as the placement destination among the plurality of arithmetic processing devices,
Controlling the first type virtual machine to be executed in preference to the second type virtual machine;
An information processing program for executing a process.

(Supplementary note 13)
From the server group, the first type of virtual machine that executes predetermined processing by bidirectional communication with other devices and routine processing that performs predetermined processing by a predetermined time without interactive input / output Select the assignment destination of the virtual machine group including the second type of virtual machine
Obtaining a predicted value of the processing unit usage rate based on the processing unit usage rate of the virtual machine group when the virtual machine group is assigned to the same server;
Determining whether the obtained predicted value is less than or equal to a threshold value based on the number of processing units of the selected server;
If it is determined that the threshold value is less than or equal to the threshold value, the virtual machine group is allocated to the selected server so that the first type virtual machine is preferentially executed over the second type virtual machine;
An assignment program for executing a process.

1701 Reception Unit 1702 Identification Unit 1703 Selection Unit 1704 Acquisition Unit 1705 Determination Unit 1706 Allocation Unit 1707 Detection Unit 1708 Control Unit 1801 Reception Unit 1802 Determination Unit 1803 Allocation Unit 1804 Control Unit

Claims (14)

A first type of virtual machine that executes a predetermined process by bidirectional communication with another device, and a second type of virtual machine that does not have interactive input / output and executes a standard process that performs the predetermined process by a predetermined time A predicted value of the processor utilization rate when the first type virtual machine becomes highly loaded due to two-way communication with the other device in the virtual machine group including the machine, An accepting unit that accepts the virtual machine group from which the predicted value of the arithmetic processing device usage rate up to the predetermined time is equal to or less than a threshold value;
A determination unit that determines an arrangement destination of the virtual machine group received by the reception unit to be any one of the plurality of arithmetic processing devices so as to be within the processing capacity of the arithmetic processing device;
An allocating unit for allocating the virtual machine group to the arithmetic processing device determined as the placement destination by the determining unit among the plurality of arithmetic processing devices;
A control unit that controls the first type virtual machine to be executed in preference to the second type virtual machine ,
The determination unit
When the first type virtual machine has already been placed in any one of the plurality of processing units, the second type virtual machine has been placed in the first type virtual machine. An information processing apparatus characterized by being determined as an arithmetic processing apparatus. A first type of virtual machine that executes a predetermined process by bidirectional communication with another device, and a second type of virtual machine that does not have interactive input / output and executes a standard process that performs the predetermined process by a predetermined time A predicted value of the processor utilization rate when the first type virtual machine becomes highly loaded due to two-way communication with the other device in the virtual machine group including the machine, An accepting unit that accepts the virtual machine group from which the predicted value of the arithmetic processing device usage rate up to the predetermined time is equal to or less than a threshold value;
A determination unit that determines an arrangement destination of the virtual machine group received by the reception unit to be any one of the plurality of arithmetic processing devices so as to be within the processing capacity of the arithmetic processing device;
An allocating unit for allocating the virtual machine group to the arithmetic processing device determined as the placement destination by the determining unit among the plurality of arithmetic processing devices;
A control unit that controls the first type virtual machine to be executed in preference to the second type virtual machine,
The determination unit
When the second type virtual machine has already been placed in any one of the plurality of processing units, the second type virtual machine has been placed as the placement destination of the first type virtual machine. An information processing apparatus characterized by being determined as an arithmetic processing apparatus. From the server group, the first type of virtual machine that executes predetermined processing by bidirectional communication with other devices and routine processing that performs predetermined processing by a predetermined time without interactive input / output A selection unit that selects an assignment destination of a virtual machine group including:
When the first type virtual machine is heavily loaded by bidirectional communication with the other device based on the processing device usage rate of the virtual machine group when the virtual machine group is assigned to the same server An acquisition unit that acquires the predicted value of the arithmetic processing device usage rate and the predicted value of the arithmetic processing device usage rate up to the predetermined time;
The predicted value of the processing device usage rate when the first type of virtual machine acquired by the acquisition unit is heavily loaded is a threshold value based on the number of processing units of the server selected by the selection unit Whether or not the predicted value of the arithmetic processing device usage rate acquired by the acquisition unit up to the predetermined time is equal to or less than a threshold value based on the number of arithmetic processing devices of the server selected by the selection unit. A determination unit for determining whether or not
The predicted value of the arithmetic processing unit usage rate when the first type virtual machine is heavily loaded by the determination unit is equal to or less than a threshold and the predicted processing unit usage rate until the predetermined time Allocating unit for allocating the virtual machine group to the selected server so that the first type of virtual machine is executed in preference to the second type of virtual machine when it is determined that the value is equal to or less than the threshold value And having
The selection unit includes:
Select the server on which the first type virtual machine has been placed,
The allocation unit is
An allocation apparatus that allocates the second type of virtual machine to the selected server. A specifying unit for specifying the second type virtual machine having the longest estimated remaining execution time from the second type virtual machine group already arranged in the server group;
The selection unit includes:
Selecting a server other than the server on which the second type virtual machine with the longest predicted remaining execution time specified by the specifying unit is placed, and on which the first type virtual machine has been placed. The allocation device according to claim 3. From the server group, the first type of virtual machine that executes predetermined processing by bidirectional communication with other devices and routine processing that performs predetermined processing by a predetermined time without interactive input / output A selection unit that selects an assignment destination of a virtual machine group including:
When the first type virtual machine is heavily loaded by bidirectional communication with the other device based on the processing device usage rate of the virtual machine group when the virtual machine group is assigned to the same server An acquisition unit that acquires the predicted value of the arithmetic processing device usage rate and the predicted value of the arithmetic processing device usage rate up to the predetermined time;
The predicted value of the processing device usage rate when the first type of virtual machine acquired by the acquisition unit is heavily loaded is a threshold value based on the number of processing units of the server selected by the selection unit Whether or not the predicted value of the arithmetic processing device usage rate acquired by the acquisition unit up to the predetermined time is equal to or less than a threshold value based on the number of arithmetic processing devices of the server selected by the selection unit. A determination unit for determining whether or not
The predicted value of the arithmetic processing unit usage rate when the first type virtual machine is heavily loaded by the determination unit is equal to or less than a threshold and the predicted processing unit usage rate until the predetermined time Allocating unit for allocating the virtual machine group to the selected server so that the first type of virtual machine is executed in preference to the second type of virtual machine when it is determined that the value is equal to or less than the threshold value And having
The selection unit includes:
Select the server on which the second type of virtual machine has been placed,
The allocation unit is
An allocation apparatus that allocates the first type virtual machine to the selected server. The allocation apparatus according to claim 3, further comprising a control unit configured to control power consumption of a server to which no virtual machine is allocated from the server group. Computer
A first type of virtual machine that executes a predetermined process by bidirectional communication with another device, and a second type of virtual machine that does not have interactive input / output and executes a standard process that performs the predetermined process by a predetermined time A predicted value of the processor utilization rate when the first type virtual machine becomes highly loaded due to two-way communication with the other device in the virtual machine group including the machine, The virtual machine group in which the predicted value of the arithmetic processing device usage rate up to a predetermined time is equal to or less than a threshold value is received from the allocation device,
The placement destination of the accepted virtual machine group is determined to be any one of the plurality of arithmetic processing devices so as to be within the processing capacity of the arithmetic processing device,
Assign the virtual machine group to the arithmetic processing device determined as the placement destination among the plurality of arithmetic processing devices,
Executing a process of controlling the first type virtual machine to be executed in preference to the second type virtual machine;
The determination process is as follows:
When the first type virtual machine has already been placed in any one of the plurality of processing units, the second type virtual machine has been placed in the first type virtual machine. An information processing method characterized by deciding to an arithmetic processing device. Computer
A first type of virtual machine that executes a predetermined process by bidirectional communication with another device, and a second type of virtual machine that does not have interactive input / output and executes a standard process that performs the predetermined process by a predetermined time A predicted value of the processor utilization rate when the first type virtual machine becomes highly loaded due to two-way communication with the other device in the virtual machine group including the machine, The virtual machine group in which the predicted value of the arithmetic processing device usage rate up to a predetermined time is equal to or less than a threshold is received from the allocation device
The placement destination of the accepted virtual machine group is determined to be any one of the plurality of arithmetic processing devices so as to be within the processing capacity of the arithmetic processing device,
Assign the virtual machine group to the arithmetic processing device determined as the placement destination among the plurality of arithmetic processing devices,
Executing a process of controlling the first type virtual machine to be executed in preference to the second type virtual machine ;
The determination process is as follows:
When the second type virtual machine has already been placed in any one of the plurality of processing units, the second type virtual machine has been placed as the placement destination of the first type virtual machine. An information processing method characterized by deciding to an arithmetic processing device . Computer
From the server group, the first type of virtual machine that executes predetermined processing by bidirectional communication with other devices and routine processing that performs predetermined processing by a predetermined time without interactive input / output Select the assignment destination of the virtual machine group including the second type of virtual machine
When the first type virtual machine is heavily loaded by bidirectional communication with the other device based on the processing device usage rate of the virtual machine group when the virtual machine group is assigned to the same server A predicted value of the arithmetic processing device usage rate and a predicted value of the arithmetic processing device usage rate up to the predetermined time,
The predicted value of the arithmetic processing unit usage rate when the acquired first type virtual machine is heavily loaded is equal to or less than a threshold value based on the number of arithmetic processing units of the selected server, and the acquired Determining whether or not the predicted value of the arithmetic processor usage rate up to a predetermined time is equal to or less than a threshold value based on the number of arithmetic processors in the selected server;
The predicted value of the processing device usage rate when the first type virtual machine is heavily loaded is less than or equal to a threshold value, and the predicted value of the processing device usage rate up to the predetermined time is less than or equal to the threshold value If it is determined that, the first type of virtual machine to preferentially execute than the second type of virtual machine, you assign the virtual machine group to the selected server, and executes the process,
The process to select is
Select the server on which the first type virtual machine has been placed,
The assigning process is:
An allocation method comprising: allocating the second type virtual machine to the selected server . Computer
From the server group, the first type of virtual machine that executes predetermined processing by bidirectional communication with other devices and routine processing that performs predetermined processing by a predetermined time without interactive input / output Select the assignment destination of the virtual machine group including the second type of virtual machine
When the first type virtual machine is heavily loaded by bidirectional communication with the other device based on the processing device usage rate of the virtual machine group when the virtual machine group is assigned to the same server A predicted value of the arithmetic processing device usage rate and a predicted value of the arithmetic processing device usage rate up to the predetermined time,
The predicted value of the arithmetic processing unit usage rate when the acquired first type virtual machine is heavily loaded is equal to or less than a threshold value based on the number of arithmetic processing units of the selected server, and the acquired Determining whether or not the predicted value of the arithmetic processor usage rate up to a predetermined time is equal to or less than a threshold value based on the number of arithmetic processors in the selected server;
The predicted value of the processing device usage rate when the first type virtual machine is heavily loaded is less than or equal to a threshold value, and the predicted value of the processing device usage rate up to the predetermined time is less than or equal to the threshold value If it is determined that the virtual machine group is assigned to the selected server so that the first type of virtual machine is executed in preference to the second type of virtual machine,
The process to select is
Select the server on which the second type of virtual machine has been placed,
The assigning process is:
An allocation method comprising: allocating the first type virtual machine to the selected server. On the computer,
A first type of virtual machine that executes a predetermined process by bidirectional communication with another device, and a second type of virtual machine that does not have interactive input / output and executes a standard process that performs the predetermined process by a predetermined time A predicted value of the processor utilization rate when the first type virtual machine becomes highly loaded due to two-way communication with the other device in the virtual machine group including the machine, The virtual machine group in which the predicted value of the arithmetic processing device usage rate up to a predetermined time is equal to or less than a threshold value is received from the allocation device,
The placement destination of the accepted virtual machine group is determined to be any one of the plurality of arithmetic processing devices so as to be within the processing capacity of the arithmetic processing device,
Assign the virtual machine group to the arithmetic processing device determined as the placement destination among the plurality of arithmetic processing devices,
Controlling the first type of virtual machine to be executed in preference to the second type of virtual machine;
The determination process is as follows:
When the first type virtual machine has already been placed in any one of the plurality of processing units, the second type virtual machine has been placed in the first type virtual machine. An information processing program that is determined to be an arithmetic processing device. On the computer,
A first type of virtual machine that executes a predetermined process by bidirectional communication with another device, and a second type of virtual machine that does not have interactive input / output and executes a standard process that performs the predetermined process by a predetermined time A predicted value of the processor utilization rate when the first type virtual machine becomes highly loaded due to two-way communication with the other device in the virtual machine group including the machine, The virtual machine group in which the predicted value of the arithmetic processing device usage rate up to a predetermined time is equal to or less than a threshold value is received from the allocation device,
The placement destination of the accepted virtual machine group is determined to be any one of the plurality of arithmetic processing devices so as to be within the processing capacity of the arithmetic processing device,
Assign the virtual machine group to the arithmetic processing device determined as the placement destination among the plurality of arithmetic processing devices,
Controlling the first type of virtual machine to be executed in preference to the second type of virtual machine;
The determination process is as follows:
When the second type virtual machine has already been placed in any one of the plurality of processing units, the second type virtual machine has been placed as the placement destination of the first type virtual machine. An information processing program that is determined to be an arithmetic processing device. On the computer,
From the server group, the first type of virtual machine that executes predetermined processing by bidirectional communication with other devices and routine processing that performs predetermined processing by a predetermined time without interactive input / output Select the assignment destination of the virtual machine group including the second type of virtual machine
When the first type virtual machine is heavily loaded by bidirectional communication with the other device based on the processing device usage rate of the virtual machine group when the virtual machine group is assigned to the same server A predicted value of the arithmetic processing device usage rate and a predicted value of the arithmetic processing device usage rate up to the predetermined time,
The predicted value of the arithmetic processing unit usage rate when the acquired first type virtual machine is heavily loaded is equal to or less than a threshold value based on the number of arithmetic processing units of the selected server, and the acquired Determining whether or not the predicted value of the arithmetic processor usage rate up to a predetermined time is equal to or less than a threshold value based on the number of arithmetic processors in the selected server;
The predicted value of the processing device usage rate when the first type virtual machine is heavily loaded is less than or equal to a threshold value, and the predicted value of the processing device usage rate up to the predetermined time is less than or equal to the threshold value If it is determined that the virtual machine group is assigned to the selected server so that the first type virtual machine is executed in preference to the second type virtual machine, the process is executed.
The process to select is
Select the server on which the first type virtual machine has been placed,
The assigning process is:
An allocation program for allocating the second type of virtual machine to the selected server. On the computer,
From the server group, the first type of virtual machine that executes predetermined processing by bidirectional communication with other devices and routine processing that performs predetermined processing by a predetermined time without interactive input / output Select the assignment destination of the virtual machine group including the second type of virtual machine
When the first type virtual machine is heavily loaded by bidirectional communication with the other device based on the processing device usage rate of the virtual machine group when the virtual machine group is assigned to the same server A predicted value of the arithmetic processing device usage rate and a predicted value of the arithmetic processing device usage rate up to the predetermined time,
The predicted value of the arithmetic processing unit usage rate when the acquired first type virtual machine is heavily loaded is equal to or less than a threshold value based on the number of arithmetic processing units of the selected server, and the acquired Determining whether or not the predicted value of the arithmetic processor usage rate up to a predetermined time is equal to or less than a threshold value based on the number of arithmetic processors in the selected server;
The predicted value of the processing device usage rate when the first type virtual machine is heavily loaded is less than or equal to a threshold value, and the predicted value of the processing device usage rate up to the predetermined time is less than or equal to the threshold value If it is determined that the virtual machine group is assigned to the selected server so that the first type virtual machine is executed in preference to the second type virtual machine, the process is executed.
The process to select is
Select the server on which the second type of virtual machine has been placed,
The assigning process is:
An allocation program for allocating the first type virtual machine to the selected server.

Images