POWER CONTROL DEVICE, POWER CONTROL METHOD, AND POWER CONTROL PROGRAM

Abstract

This power supply control device (2) has: a supply power prediction unit (21) for determining the excess and insufficiency of a power supply state in accordance with the status of power supply and demand to each server installed in a data center (3); and a power control unit (22) for changing the total number of operating servers in the data center (3) and at least one of the power settings of the respective servers in operation so as to eliminate the excess and insufficiency of the power supply state. The power control unit (22) further changes the allocation of service programs to the servers on the basis of the power supply state, said service programs operating on the respective servers.

Description

[Technical field]

[0001] The present invention relates to electric power Control device, electric power A control method, and electric power Regarding the control program. [Background technology]

[0002] A data center is equipped with multiple servers, and applications deployed on each server consume power when they are executed. When using renewable energy as a power generation facility to supply power to a data center, the power supply may be unstable, and some servers may experience power shortages. Therefore, Non-Patent Document 1 describes a method of operating servers by distributing the load among geographically distributed data centers. [Prior art documents] [Non-patent literature]

[0003] [Non-Patent Document 1] H. Goudarzi et al, “Geographical Load Balancing for Online Service Applications in Distributed Datacenters,” Published in: 2013 IEEE Sixth International Conference on Cloud Computing, [online], IEEE, [Retrieved December 25, 2020], Internet〈URL:https: / / ieeexplore.ieee.org / abstract / document / 6676714〉 Summary of the Invention [Problem to be solved by the invention]

[0004] On the other hand, if it is not possible to utilize geographically distributed data centers as described in Non-Patent Document 1, it will be necessary to either use regular power supplies such as thermal power plants in addition, or to prepare sufficient power storage facilities. Therefore, when considering renewable energy as the main focus, power generation methods that have a large bias in the times when power generation is possible require very large storage facilities to stabilize the power supply, and if sufficient storage cannot be secured, energy will be wasted.

[0005] Therefore, a main object of the present invention is to operate a server stably while effectively utilizing the supplied power even in an environment where the power supply is unstable. [Means for solving the problem]

[0006] In order to solve the above problems, the power control device of the present invention has the following features. The present invention includes a determination unit that determines whether a power supply state is excessive or insufficient in accordance with a power supply and demand state of each server installed in a data center; a control unit that changes at least one of a total number of servers operating in the data center, a power setting of each server in operation, and an arrangement of a service program operating on each server to the server, so as to eliminate an excess or deficiency in the power supply state, When the power supply state is an excess power state, the control unit executes at least one of a first control of increasing the total number of servers operating in the data center, a second control of increasing power consumption as a power setting of each server in operation, a third control of relocating the service program operating on a predetermined server to a server consuming more power than the predetermined server, and a fourth control of increasing the number of the service programs operating on a server. year, the control unit determines whether each of the first control, the second control, the third control, and the fourth control is executable, and when there are a plurality of executable controls, executes the controls in an order according to a predetermined policy until the excess power state is resolved. It is characterized by: Effect of the Invention

[0007] According to the present invention, even in an environment where the power supply is unstable, it is possible to operate a server stably while effectively utilizing the supplied power. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a configuration diagram of a data center system according to an embodiment of the present invention. [Diagram 2] FIG. 4 is a state transition diagram relating to power control according to the present embodiment. [Diagram 3] 1 is a hardware configuration diagram of a power control device according to an embodiment of the present invention; [Figure 4] 1 is a table showing the state before power control for each server constituting a data center according to the present embodiment. [Diagram 5] This is the state after power control for the table of FIG. 4 according to this embodiment. [Figure 6] 3 is a flowchart showing an outline of processing performed by the power control device according to the present embodiment. [Figure 7] 10 is a flowchart showing details of a process for distributing requests while saving power according to the present embodiment; [Figure 8] 5 is a flowchart showing the first half of a power saving control process according to the present embodiment. [Figure 9] 10 is a flowchart showing the second half of the power saving control process according to the embodiment. [Figure 10] 4 is a flowchart showing a performance control process according to the present embodiment. [Figure 11] 4 is a table showing conditions for executing each action of power control according to the present embodiment. [Figure 12] 12 is a table showing examples in which each action of the process in FIG. 7 is executed in accordance with the table in FIG. 11 according to this embodiment. [Figure 13] 4 is a table showing examples of execution of each action in the power saving control process according to the present embodiment. [Figure 14] 11 is a table showing examples of execution of each action in the performance control process according to the present embodiment. [Figure 15] 1 is a graph showing the relationship between power consumption and performance according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, the present embodiment will be described with reference to the drawings.

[0010] FIG. 1 is a configuration diagram of a data center system 9. In a data center system 9, a power generation facility 1, a power control device 2, and a data center 3 are connected via a network, as indicated by thin arrows. Furthermore, the power generation facility 1 supplies power to the data center 3, as indicated by thick arrows. The power generation facility 1 may be a conventional power generation facility such as a thermal power generation facility whose power supply amount is controllable and stable, or may be a renewable energy power generation facility such as a solar power generation facility whose power supply amount is uncontrollable due to factors such as the weather. In the following, the present specification will be exemplified by the case of solar power generation.

[0011] The data center 3 is equipped with one or more servers, and each server provides various services by executing programs installed from outside. The service program executed on the server may be an application that can be executed directly on the physical resources (bare metal) of the server, or it may be middleware that provides an operating environment for the application as an intermediate layer between the server and the application. The middleware is, for example, a virtual machine (VM) and a container. In the following, the case of a VM will be illustrated.

[0012] The power control device 2 includes a supply power prediction unit (determination unit) 21 and a power control unit (control unit) 22. The power supply prediction unit 21 predicts the power supply from the power generation facility 1 to the data center 3 on a time basis (such as every hour). For example, if the power generation facility 1 is a solar power generation facility, the power supply is greatly affected by the amount of solar radiation. Thus, the power supply prediction unit 21 receives weather forecast data such as "It's sunny now, but the weather forecast predicts rain in an hour" and predicts future power supply, such as "The amount of power supply will be 50% less than now."

[0013] The supply power prediction unit 21 predicts the surplus or shortage of power supplied from the power generation facility 1 to the data center 3 for each unit of time (such as every hour). Therefore, the supply power prediction unit 21 finds the surplus or shortage of power supplied by comparing the predicted value of the supply power with the actual measured value of power consumption obtained by monitoring the power consumption of the servers operating in the data center 3. Alternatively, the supply power prediction unit 21 may receive input data from an external system that specifies the surplus or shortage of supply power for each time unit (for example, for each hour).

[0014] FIG. 2 is a state transition diagram relating to power control. The power supply state 70 indicates a power surplus 71 or a power shortage 72 as a state of excess or shortage of power supplied from the power generation facility 1 to the data center 3 . The power control unit 22 executes power control for each server in the data center 3 in accordance with the power supply state 70, which is output data from the power supply prediction unit 21, so as to eliminate the surplus or deficiency.

[0015] The server setting state 80 is a power state defined for each server in the data center 3, and state transitions are possible according to BIOS (Basic Input / Output System) settings such as CPU settings. Performance 81 is a server setting that is less power efficient (performance / watt) but consumes more power to maximize performance. Balanced 82 is also a setting that improves power efficiency by slightly reducing power consumption and performance compared to Performance 81. The Minimum 83 will be described as a setting that does not transition to other settings, for example. The Minimum 83 has even lower power consumption and performance than the Balance 82, and instead improves power efficiency.

[0016] Therefore, the power consumption performance ratio of a server varies depending on the design characteristics of the CPU it uses, as shown in the following example. By operating a server with a high-performance CPU design that prioritizes maximum performance (hereinafter referred to as a "high-performance server") at performance 81, power consumption and maximum performance can be increased, but power efficiency will be reduced. By operating a server with a low-power CPU design that prioritizes power efficiency (hereinafter referred to as a "low-power server") at minimum 83, power consumption and performance are low, but power efficiency can be increased. In this way, the data center 3 may be constructed as a heterogeneous environment in which a plurality of types of server sets with different computer properties are mixed. The computer properties are defined, for example, by the design characteristics of the CPU that determine the performance and power efficiency. On the other hand, the data center 3 may be constructed by a set of servers with the same properties.

[0017] Returning to Fig. 1, the power control unit 22 executes at least one of the power controls exemplified below as first to fourth controls in accordance with the power supply state 70. This makes it possible to make maximum use of the power that can be supplied while balancing the power demand. [First control] Change the total number of servers in operation. If there is a power shortage 72, reduce the number of servers in operation, and if there is a power surplus 71, increase the number of servers in operation. [Second control] Change the performance settings of each operating server. For example, if there is a power shortage 72, change performance 81 to balance 82, and if there is excess power 71, change balance 82 to performance 81. Alternatively, the type of server may be switched without changing the total number of operating servers. For example, if there is a power shortage 72, more low-power servers are operated, and if there is excess power 71, more high-performance servers are operated. [Third control] Change the placement destination of the VM (see Fig. 4 for details). For example, if there is a power shortage 72, the VM is relocated to a server with good power efficiency, and if there is a power surplus 71, the VM is relocated to a server with high power consumption. · [Fourth control] Change the running / non-running status of VMs running on the server (details in Fig. 4). If there is a power shortage 72, the number of running VMs is reduced, and if there is a power surplus 71, the number of running VMs is increased.

[0018] FIG. 3 is a diagram showing the hardware configuration of the power control device 2. As shown in FIG. The power control device 2 is configured as a computer 900 having a CPU 901 , a RAM 902 , a ROM 903 , a HDD 904 , a communication I / F 905 , an input / output I / F 906 , and a media I / F 907 . The communication I / F 905 is connected to an external communication device 915. The input / output I / F 906 is connected to an input / output device 916. The media I / F 907 reads and writes data from a recording medium 917. Furthermore, the CPU 901 executes the programs (supply power prediction unit 21, power control unit 22, and programs included in the power control device 2) loaded into the RAM 902. electric power The program controls each processing unit by executing a program (such as a control program). This program can be distributed via a communication line or recorded on a recording medium 917 such as a CD-ROM and distributed.

[0019] FIG. 4 is a table showing the state of each server constituting the data center 3 before power control. Table 101 shows the state of the hardware in data center 3 before power control. Data center 3 is equipped with four high-performance servers (SA1 to SA4) and five low-power servers (SB1 to SB5), of which six are powered on and in operation. Even in the same power-on state, the high-performance server with a performance setting of 81 consumes 300 [W] of power, which is more than the 200 [W] of the high-performance server with a balance setting of 82.

[0020] Table 102 shows the state of the software in data center 3 before power control. For example, two VMs, V01 and V02, are placed on the server SA1. Each VM is assigned a request label. For example, the request label [150:variable 100~] of V01 indicates that the required power is 150 [W] and that the required performance may be variable within a range of 100 [W] or more. The request label of a VM indicates a configuration parameter when requesting the creation of a VM to a VM orchestrator (Kubernetes / OpenStack, etc.). Below, a performance variation label and a priority label are explained as examples of request labels.

[0021] The performance variation label is a label that specifies whether or not the variation of the required performance of the VM is allowed. The required performance is defined by the required number of CPU cores, the required memory usage, the required GPU specifications (model number), etc. The performance variation label "fixed" is specified when variation in the required performance is not permitted and the same performance is constantly required as a service level agreement (SLA). The performance variation label "variable" is specified when required performance is allowed to vary, and there is a lower limit to the required performance, but performance can also be allowed to increase. For example, "variable" is specified in cases where a service can be established even if the video processing performance (resolution, etc.) changes, such as video distribution, or in the case of batch processing where a certain total amount of processing is performed but the execution time is not important.

[0022] The priority label is referenced to determine whether a VM among multiple types of VMs should be given priority over other VMs across the entire system, and whether a VM should be allocated more server performance and run with priority over other VMs. The priority label can be any integer between 0 and 3, and the higher the number, the higher the priority it has over other VMs. The priority label "0" allows for periods of time when the VM is stopped and not in operation (in Figure 4, SB1's "Shutdown Possible"), and is assigned when the VM is not required to be running constantly, such as in batch processing.

[0023] Fig. 5 shows a state after power control for the table of Fig. 4. Table 103 illustrates a case where the power shortage 72 of Fig. 4 is resolved in Fig. 5. The power control unit 22 saves power by reducing the number of running VMs. To do this, the power control unit 22 shuts down V07 on the server SB1 and then turns off the power of the server SB1. This saves 100 [W] of power. Furthermore, since the performance change labels of V01 and V02 are both "variable," the power control unit 22 changes the performance setting of the server SA1 from performance 81 to balance 82, thereby saving 100 [W] of power.

[0024] FIG. 6 is a flowchart showing an outline of the processing performed by the power control device 2. As shown in FIG. The power control unit 22 determines whether or not a request to deploy a VM on a server has been received from an external system (S11). If Yes in S11, the process proceeds to S12, and if No, the process proceeds to S21. The power supply prediction unit 21 branches the process based on the current or near future power supply state 70 (S12). If there is a power surplus 71 in S12, the process proceeds to S14, and if there is a power shortage 72, the process proceeds to S13.

[0025] The power control unit 22 deploys the VM of the request received in S11 to a server in the data center 3 (S14). Here, the selection criteria for the server to which the VM is to be deployed are as follows. (Criterion 1) The server must be capable of running the required performance stated in the VM's request label. For example, if a VM specifies a required performance of 70 [W], a server with a power consumption margin of 70 [W] or more will be selected. For example, assume that the total power consumption of Server A is 500 [W], and the power consumption of another VM currently running is 100 + 200 + 150 = a total of 450 [W]. In this case, Server A can consume the remaining 50 [W], but is not selected due to insufficient margin for the required performance of 70 [W].

[0026] (Criterion 2) Among the set of servers that satisfy criterion 1, a server is selected so that the performance fluctuation labels of the VMs already in operation and the performance fluctuation labels of the VMs of the request received this time in S11 are as similar as possible (so that the ratio of identical labels is increased). This makes it easier to change the performance settings of the server, and increases the effect of the change, by matching the performance fluctuation labels of the set of VMs running on the same server, as in the case of server SA1 in Figures 4 and 5. Note that (Criterion 1) is mandatory, but (Criterion 2) is optional.

[0027] On the other hand, the process of deploying the request while saving power (S13) is basically the same as S14, but currently there is no server that satisfies (criterion 1) due to a power shortage 72. Therefore, before deploying the VM of the request, control is performed to save power in the entire data center 3 (details are shown in Fig. 7). When there is no requesting VM, the power supply prediction unit 21 branches the process based on the power supply state 70, as in S12 (S21). If there is a power surplus 71 in S21, the process proceeds to S23 where performance control is performed to increase power consumption (see FIG. 10 for details), and if there is a power shortage 72, the process proceeds to S22 where power saving control is performed to reduce power consumption (see FIG. 8 and FIG. 9 for details).

[0028] Fig. 7 is a flowchart showing the details of the process (S13) of deploying a request while saving power. When a VM specifies a required performance of 70 [W], the process in Fig. 7 ensures that the server has a spare capacity of 70 [W] before deploying the VM. Regarding the first letter of the symbols used in the flowcharts below, a process beginning with "C" indicates a condition (C for condition), and a process beginning with "A" indicates power control (A for action).

[0029] The power control unit 22 determines whether or not there is a VM whose performance variation label is "variable" (C11), and determines whether or not there is a server whose performance setting is performance 81 (C12). If (Yes in C11) and (Yes in C12), the power control unit 22 saves power by changing the setting of the server whose performance setting is performance 81 to balance 82 (A13). In addition, when there are multiple candidates for the server whose settings are to be changed in A13, the settings may be changed in order of the lowest total value of the priority labels of the running VMs. Alternatively, if the total values ​​of the priority labels are the same, a server may be selected at random and its settings may be changed.

[0030] The power control unit 22 judges whether or not the power consumption of the already deployed VM can be reduced at another deployment destination than the current deployment destination and whether or not there is spare capacity to deploy the VM at the other deployment destination (C13). If (Yes in C11) and (No in C12) and (Yes in C13), the power control unit 22 saves power by relocating the VM to another deployment destination that has better power efficiency than the current deployment destination (A14). Note that when there are multiple candidates for the VM to be relocated in A14, the VM is selected at random, for example.

[0031] On the other hand, if the answer is (No in C11) or (No in C13), the process proceeds to C14. The power control unit 22 determines whether or not there is a VM with a label (priority label=0) that can be shut down (stopped) (C14). The power control unit 22 randomly selects one target VM for which the result of C14 is Yes, and shuts it down to reduce power consumption (A11). In the case of (No in C14), if a server that satisfies the required performance of the requested VM exists, the power control unit 22 may deploy the VM, or if the server does not exist, may discard the request (A12).

[0032] FIG. 8 is a flowchart showing the first half of the power saving control process (S22). 9 is a flow chart showing the second half of the power saving control process (S22). The processes C11 to C14 starting with "C" and the processes A11 to A14 starting with "A" used in FIGS. 8 and 9 are as described in FIG.

[0033] The power control unit 22 determines whether the power shortage 72 has been resolved (S101). If Yes in S101, the process of S22 ends, and if No, the process proceeds to S102. When there is an empty server on which no VM is running (Yes in S102), the power control unit 22 saves power by powering off the empty server (A15). Note that when there are multiple candidates for the empty server in A15, for example, the power control unit 22 powers off the empty server with the highest power consumption.

[0034] The power control unit 22 proceeds to C11 if the determination in C11 is No, proceeds to C12 if the determination in C11 is Yes, and starts process A13 as in Fig. 7 if the determination in C12 is Yes. If the determination in C11 is No, proceeds to S103. When it is possible to consolidate the VMs to another server by relocating the running VMs to the other server (Yes in S103), the power control unit 22 consolidates the VMs to the other server (A16). As a result, the server from which the VMs are to be relocated becomes an available server, and can be targeted for power off in the next S102. Note that it is preferable that the server from which the VMs are to be relocated is a server with a small number of VMs allocated thereto. If the answer is No in S103, the process proceeds from connector A to C14 in FIG. 9, and if the answer is No in C12, the process proceeds from connector B to C13 in FIG.

[0035] 9, and performs processing of A14 if C13 is Yes. The power control unit 22 performs processing of C14 if the result is the connector A or (No in C13), and performs processing of A11 if the result is Yes in C14. If the answer is No in C14, the process A12 is performed in FIG. 7, but in FIG. 9, this is replaced by the process A12B. The power control unit 22 conserves power by sequentially shutting down (stopping) the running VMs in order of lowest priority label and lowest operating status such as CPU usage until the power shortage 72 is resolved (A12B).

[0036] FIG. 10 is a flowchart showing the performance control process (S23). 7 to 9 have described control for saving power to resolve power shortage 72, but since there is a power surplus 71 in FIG. 10, control is performed to increase power consumption and improve the processing performance of the VM.

[0037] The power control unit 22 determines whether or not there is a VM whose performance variation label is "variable" (C11), and determines whether or not there is a server whose performance setting is balanced 82 (C12C). If (Yes in C11) and (Yes in C12C), the power control unit 22 increases power consumption by changing the setting of the server whose performance setting is balanced 82 to performance 81 (A13C).

[0038] If (No in C12C), the power control unit 22 determines whether or not the power consumption of the already deployed VM can be increased at another deployment destination than the current deployment destination and whether there is spare capacity to deploy the VM at the other deployment destination (C13C). If (Yes in C11) and (No in C12C) and (Yes in C13C), the power control unit 22 improves the performance of the VM by relocating the VM to another deployment destination that consumes more power than the current deployment destination (A14C).

[0039] On the other hand, if the answer is (No in C11) or (No in C13C), the process proceeds to C14C. The power control unit 22 determines whether or not there is a server that is currently stopped but can be started within the range of surplus power in the data center 3 (C14C). The power control unit 22 randomly selects one of the servers for which the answer is Yes in C14, and starts it up, thereby increasing power consumption (A11C). Furthermore, the power control unit 22 may start up a new VM on the new server when the new server is started up. If the answer is No in C14C, the process of S23 is terminated. As described above, each of the power saving actions A11 to A14 and each of the power increasing actions A11C to A14C are executed when a predetermined condition such as C11 is satisfied. The relationship between the conditions and the actions will be summarized below with reference to Figs. 11 to 14.

[0040] FIG. 11 is a table showing the conditions for executing each action of power control. The row in the table with the status "Deployment" and power "Insufficient" corresponds to the process (S13) in FIG. The row in the table with the status "control" and power "insufficient" corresponds to the process (S22) in FIGS. The row in the table with the status "under control" and power "excess" corresponds to the process (S23) in FIG.

[0041] For example, the following two cases are shown as examples of conditions for executing the action A11 in FIG. If (C11: No) and (C14: Yes) (Yes in C11) and (No in C12) and (No in C13) and (Yes in C14) 11, the only condition for executing action A11 is that C14 must be satisfied (Yes in C14). In other words, regardless of the results of C11, C12, and C13, action A11 will be executed if (Yes in C14).

[0042] FIG. 12 is a table showing examples in which each action of the process (S13) in FIG. 7 is executed in accordance with the table in FIG. This table associates, for each of rows L11 to L17 indicated by ID, individual judgment results as to whether or not each condition is satisfied, with actions that can be derived from those judgment results using the table in Fig. 11. Note that blank cells for conditions indicate that it has not been judged whether or not the condition is satisfied. For example, in row L11, (C11 is not satisfied) and (C14 is satisfied), so action A11 is executed according to the rule in the first row of the table in FIG. On the other hand, in row L17, all of the conditions C11 to C14 are satisfied, so actions A11, A13, and A14 are executed according to the rules in the first, third, and fourth rows of the table in Fig. 11. In other words, an example that satisfies action A11 may be row L11 or row L17.

[0043] FIG. 13 is a table showing examples of execution of each action in the power saving control process (S22). 12 and 13 are common in that they involve power saving control in response to a power shortage 72, so the actions in FIG. 12 and the actions in FIG. 13 are started under substantially the same conditions. On the other hand, in FIG. 13, since there is no request, the action A12 of discarding the request is replaced with the action A12B of stopping the running VM.

[0044] FIG. 14 is a table showing examples of when each action is executed in the performance control process (S23). For example, in row L31, (C11 is not satisfied) and (C14C is satisfied), so action A11C is executed according to the rule in the 9th row of the table in FIG. On the other hand, in row L35, all of the conditions C11 to C14C are satisfied, so actions A11C, A13C, and A14C are executed in accordance with the rules in rows 9 to 11 of the table in FIG. In this way, various actions are executed depending on the combination of conditions according to the rules in the table in FIG.

[0045] 15 is a graph showing the relationship between power consumption and performance. The horizontal axis of the graph indicates the performance provided to the VM, and the vertical axis of the graph indicates the power consumption of the VM. Each of the three curves corresponds to a server setting state 80. Curve 201 represents the minimum 83 state, curve 202 represents the balance 82 state, and curve 203 represents the performance 81 state. The figure also shows three types of VMs running on the server. The first VM only requires low required performance P1, so it can run in minimum 83, balance 82, or performance 81. The power consumption required to provide the same performance is lowest in minimum 83, followed by balance 82, and highest in performance 81.

[0046] Therefore, it is desirable to run the first VM on a server with the best power efficiency, Minimum 83. For example, in the case where the first VM is running on a server with Balance 82 in the judgment C13 of Fig. 7, the result is "Yes in C13". Therefore, the power control unit 22 can save power by relocating the first VM to a server with Minimum 83 (A14), while providing the first VM with the same performance before and after the relocation.

[0047] On the other hand, the second VM has a medium required performance P2, and therefore cannot operate at minimum 83 due to insufficient performance. Therefore, the second VM is placed on a server in balance 82 or performance 81, with the server in balance 82 being more power efficient. Note that in the server in balance 82, the lower the CPU usage rate, the lower the CPU clock and the lower the power consumption.

[0048] Furthermore, since the third VM has high required performance P3, it cannot operate due to insufficient performance even in minimum 83 or balance 82. Therefore, the placement destination of the third VM is limited to a server with performance 81. Therefore, when there is excess power 71, the number of servers with performance 81 is increased by action A13C in Fig. 10, and by operating the third VM, it is possible to increase satisfaction with the provided service. In this way, the power control unit 22 flexibly uses a policy that prioritizes power efficiency and a policy that prioritizes absolute performance depending on the state of power supply and demand.

[0049] In addition, when there are multiple options for actions to perform power control, such as in the case of row L17 in FIG. 12, the power control unit 22 may determine the order of actions to be executed in accordance with any of the policies exemplified below. (Policy 1) In order of lightest cost (load) of the action. For example, A15 → A13 → A11 → A12 → A14, A16. This allows the power reduction to be completed in a short time, although the power reduction is small. (Policy 2) In ascending order of the impact on services (for example, service outage time). For example, the order is the first control, the second control, the third control, and the fourth control, and the order is A15 → A13 → A14, A16 → A11 → A12. This makes it possible to minimize the impact on running services. This policy 2 is a general operational policy.

[0050] (Policy 3) In order of greatest power saving effect. For example, A15 → A13, A14, A16 → A11, A12. This allows for large power reductions in a short period of time. Policy 3 is used in situations where power reduction is the top priority, such as when the available power supply drops suddenly due to a disaster and cannot keep up with demand, or when system-wide problems are expected. The priority of A16 varies depending on the number of relocation targets.

[0051] Here, the order of "A→B, C→D" described in each policy is as follows: · Action A is performed with the highest priority (the symbol "A → B" means that A takes priority over B). If the power shortage is resolved as a result, the execution of actions B, C, and D is omitted. If the power shortage is not resolved even with action A, execute action B or action C with the same priority (the symbols "B,C" mean that B and C have the same priority). If the power shortage is still not resolved, perform action D.

[0052] [effect] The power control device 2 of the present invention includes a power supply prediction unit 21 that judges whether the power supply state 70 is excessive or insufficient according to the power supply and demand situation of each server installed in the data center 3; The data center 3 is characterized by having a power control unit 22 that changes at least one of the total number of servers operating in the data center 3 and the power settings of each operating server so as to eliminate any surplus or deficiency in the power supply state 70.

[0053] As a result, even in an environment where the power supply is unstable, each server in the data center 3 can effectively use the power supplied. Furthermore, each server in the data center 3 operates stably without experiencing power shortages.

[0054] The present invention is characterized in that the power control unit 22 also changes the allocation of service programs operating on each server to the server based on the power supply state 70.

[0055] This makes it possible to provide services without increasing or decreasing the number of service programs in operation, even in an environment where the power supply is unstable.

[0056] The present invention is characterized in that, when the power supply state 70 is an excess power 71, the power control unit 22 executes at least one of a first control of increasing the total number of operating servers in the data center 3, a second control of increasing the power consumption as a power setting of each operating server, a third control of relocating a service program operating on a specified server to a server consuming more power than the specified server, and a fourth control of increasing the number of service programs operating on the server.

[0057] This makes it possible to effectively utilize surplus power and improve the performance of the service program.

[0058] The present invention is characterized in that, when the power supply status 70 is a power shortage 72, the power control unit 22 executes at least one of a first control of reducing the total number of operating servers in the data center 3, a second control of reducing power consumption as a power setting of each operating server, a third control of relocating a service program operating on a specified server to a server that is more power efficient than the specified server, and a fourth control of reducing the service programs operating on the server.

[0059] This allows power to be saved and service programs to run more stably.

[0060] The present invention is characterized in that the power control unit 22 executes the control in the order of the first control, the second control, the third control, and the fourth control until the power shortage 72 is resolved.

[0061] As a result, power control is performed in the order of decreasing impact on services (for example, service outage time), thereby minimizing the impact on running services. [Explanation of symbols]

[0062] 1. Power generation facilities 2 Power control device 3 Data Center 9 Data Center Systems 21 Power supply prediction unit (judgment unit) 22 Power control section (control section) 70 Power supply status 71 Excess power (excess power condition) 72 Power shortage (power shortage state) 80 Server setting status 81 Performance 82 Balance 83 Minimum

Claims

1. a determination unit that determines whether a power supply state is excessive or insufficient according to a power supply and demand state of each server deployed in the data center; a control unit that changes at least one of a total number of servers operating in the data center, a power setting of each server in operation, and an arrangement of a service program operating on each server to the server, so as to eliminate an excess or deficiency in the power supply state, When the power supply state is an excess power state, the control unit executes at least one of a first control of increasing a total number of servers operating in the data center, a second control of increasing power consumption as a power setting of each server in operation, a third control of relocating the service program operating on a specified server to a server consuming more power than the specified server, and a fourth control of increasing the number of the service programs operating on a server, The control unit determines whether each of the first control, the second control, the third control, and the fourth control is executable, and when there are a plurality of executable controls, the control unit executes the controls in an order according to a predetermined policy until the excess power state is resolved. Power control device.

2. a determination unit that determines whether a power supply state is excessive or insufficient according to a power supply and demand state of each server deployed in the data center; a control unit that changes at least one of a total number of servers operating in the data center, a power setting of each server in operation, and an arrangement of a service program operating on each server to the server, so as to eliminate an excess or deficiency in the power supply state, when the power supply state is a power shortage state, the control unit executes at least one of a first control of reducing a total number of servers operating in the data center, a second control of reducing power consumption as a power setting of each server in operation, a third control of relocating the service program operating on a specified server to a server having better power efficiency than the specified server, and a fourth control of reducing the service program operating on a server; The control unit determines whether each of the first control, the second control, the third control, and the fourth control can be executed, and when there are a plurality of executable controls, executes the controls in an order according to a predetermined policy until the power shortage state is resolved. Power control device.

3. The control unit executes control in the order of lightest control load according to the predetermined policy. The power control device according to claim 1 or 2.

4. The control unit is characterized in that, as an order according to the predetermined policy, it executes control in the order of least impact on the service program currently in operation. The power control device according to claim 1 or 2.

5. The control unit executes control in the order of greatest power saving effect according to the predetermined policy. The power control device according to claim 2 .

6. The power control device includes a determination unit and a control unit. The determination unit determines whether a power supply state is excessive or insufficient in accordance with a power supply and demand state of each server deployed in the data center; the control unit, as a process of changing at least one of the total number of servers operating in the data center, the power settings of each operating server, and the allocation of service programs operating on each server to the servers so as to eliminate an excess or shortage in the power supply state, executes at least one of a first control of increasing the total number of servers operating in the data center, a second control of increasing power consumption as the power settings of each operating server, a third control of relocating the service programs operating on a specified server to a server consuming more power than the specified server, and a fourth control of increasing the number of service programs operating on a server, when the power supply state is an excess power state; The control unit determines whether each of the first control, the second control, the third control, and the fourth control is executable, and when there are a plurality of executable controls, the control unit executes the controls in an order according to a predetermined policy until the excess power state is resolved. Power control methods.

7. The power control device includes a determination unit and a control unit. The determination unit determines whether a power supply state is excessive or insufficient in accordance with a power supply and demand state of each server deployed in the data center; the control unit executes, when the power supply state is a power shortage state, at least one of a first control of reducing the total number of servers operating in the data center, a second control of reducing power consumption as a power setting of each operating server, a third control of relocating the service program operating on a specified server to a server having better power efficiency than the specified server, and a fourth control of reducing the service program operating on a server, as a process of changing at least one of the total number of servers operating in the data center, the power setting of each operating server, and the allocation of the service program to the server so as to eliminate an excess or shortage in the power supply state; The control unit determines whether each of the first control, the second control, the third control, and the fourth control can be executed, and when there are a plurality of executable controls, executes the controls in an order according to a predetermined policy until the power shortage state is resolved. Power control methods.

8. The control unit is characterized in that, as an order according to the predetermined policy, the control is executed in the order of lightest control load. The power control method according to claim 6 or 7.

9. The control unit executes control in the order according to the predetermined policy in the order of least impact on the service program in operation. The power control method according to claim 6 or 7.

10. The control unit executes control in the order of greatest power saving effect according to the predetermined policy. The power control method of claim 7.

11. A power control program for causing a computer to function as the power control device according to any one of claims 1 to 5.

Images