Power efficiency calculation device, power efficiency calculation method, power efficiency calculation system, and program

Abstract

Provided is a power efficiency calculation device (1) comprising: a metrics collection unit (101) that collects metrics from a physical server group (30); a metrics measurement interval determination unit (102) that extracts changes in task execution volume as frequency components at predetermined time intervals, and determines a metrics measurement interval; a power efficiency calculation interval determination unit (103) that determines, as a power efficiency calculation interval, an interval that is equal to or longer than the metrics measurement interval and is equal to or smaller than a minimum executable interval (121); and a power efficiency calculation unit (104) that calculates a power efficiency at the determined power efficiency calculation intervals using the metrics collected at the determined metrics measurement intervals.

Description

[Technical Field]

[0001] The present invention relates to a power efficiency calculation device, a power efficiency calculation method, a power efficiency calculation system, and a program for calculating power efficiency for application processing. [Background technology]

[0002] To date, various definitions have been proposed for data center power efficiency indicators. Among them, the only indicator that can measure and calculate in real time how power-efficiently a specific application is being operated is DCeP (Data Center energy Productivity) (see Non-Patent Document 1). DCeP is defined by the following formula (1):

[0003]

number

[0004] DCeP (Power Efficiency) is defined as "Useful Work Produced" (referred to herein as "Work Output") divided by "Total Energy Consumed to Perform that Work" (referred to herein as "Power Consumption"). In addition, "Useful Work Produced" is defined by the following formula (2).

[0005]

number

[0006] where "M" is the number of tasks started during the evaluation window, "V i ” is the normalization coefficient obtained by summing the task values, “U i“(t,T)” is the time-based utility function for each task, “t” is the elapsed time from the start to the completion of the task, and “T” is the absolute time of task completion. Non-Patent Document 1 proposes setting a weighting factor for tasks that are the target of "Useful Work Produced" (amount of work output) and adjusting the balance of values ​​between tasks. However, there is no mention of how this should be adjusted specifically. Furthermore, with regard to the assessment window, which is the measurement interval, it is only stated that "an interval of 20 or more task executions is desirable."

[0007] Furthermore, Non-Patent Document 2 defines "Useful Work Produced" (amount of work produced) as the energy consumed when various applications are processed in an HPC (High Performance Computing) data center, and describes the results of calculating DCeP (power efficiency). However, Non-Patent Document 2 treats all tasks equally without considering the weight of each application type or the importance of processing between tasks. In addition, the calculation interval for DCeP (power efficiency) is set to one hour.

[0008] In Non-Patent Document 3, "Useful Work Produced" is defined as the number of times two types of applications are executed within a certain period of time (Useful Computational Units), and the number of times each type of application is executed is normalized using a weighting factor (1:0.08). The calculation interval for DCeP (power efficiency) is determined to be the time it takes for a specific application to complete execution.

[0009] Here, it is assumed that the actual measured values ​​of "Useful Work Produced" (amount of work output) and "Total Energy Consumed to Perform that Work" (amount of power consumption) shown in formula (1) are obtained by using metrics collection software (resource monitoring software) to obtain metrics, which are evaluation indicators of performance, etc., from the system ("physical server group 30" described later). Prometheus, one of the de facto standards for metrics collection software, sets the default value (scrape interval) for metrics collection to one minute (see Non-Patent Document 4). With Prometheus, users must set the measurement interval appropriately according to their needs, but there are no clear guidelines on what value should be set. [Prior art documents] [Non-patent literature]

[0010] [Non-Patent Document 1] D. Anderson, et al., “A framework for data center energy productivity.” The Green Grid, 2008. [Non-patent document 2] A. Grishina, et al., “DC energy data measurement and analysis for productivity and waste energy assessment,” 2018 IEEE International Conference on Computational Science and Engineering (CSE), IEEE, 2018. [Non-patent document 3] Landon H. Sego, et al., “Implementing the data center energy productivity metric,” ACM Journal on Emerging Technologies in Computing Systems (JETC) 8.4 (2012): 1-22. [Non-patent document 4] Prometheus Authors 2014-2022, “Prometheus,” The Linux Foundation, 2022, [online], [Retrieved May 9, 2020], Internet<URL:https: / / prometheus.io / docs / prometheus / latest / configuration / configuration / > Summary of the Invention [Problem to be solved by the invention]

[0011] When calculating DCeP (power efficiency), if the frequency of measuring the metrics required for calculating DCeP (power efficiency) and the frequency of calculating DCeP (power efficiency) are not set appropriately, the following problems will occur. If metrics are measured too frequently and DCeP (power efficiency) is calculated too frequently, the power efficiency value will be affected by small changes and become unstable, and the measurements and calculations will result in wasted power consumption and storage capacity pressure. Conversely, if metrics are measured too infrequently and DCeP (power efficiency) is calculated too infrequently, changes in demand fluctuations cannot be observed.

[0012] The diagram indicated by the reference numeral 51 in Figure 5 shows an example in which the frequency of metrics measurement and the frequency of DCeP (power efficiency) calculation are appropriately set. The horizontal axis represents time [h], and the vertical axis shown as a histogram represents task processing volume [au] such as data transfer volume or power [W]. The vertical axis shown as a line graph represents power efficiency [au], and indicates that DCeP (power efficiency) calculation was performed six times (indicated by circles) within a given period.

[0013] As shown by reference numeral 51 in Fig. 5, even in the case of an application with large fluctuations in demand, if the measurement frequency of metrics is too low, the task processing volume and power consumption will be smoothed, and the changes will be overlooked, as shown by reference numeral 52 in Fig. 5. Also, as shown by reference numeral 53 in Fig. 5, if the calculation frequency of DCeP (power efficiency) is too low (three times in a given period), the sensitivity to changes in power efficiency will be low, and control opportunities will be missed.

[0014] The present invention was made in consideration of these points, and its objective is to appropriately determine the metrics measurement interval and the power efficiency calculation interval in accordance with fluctuations in demand related to application usage. [Means for solving the problem]

[0015] The power efficiency calculation device of the present invention is a power efficiency calculation device that calculates the power efficiency by executing an application installed on a group of physical servers, and is characterized by comprising: a metrics collection unit that collects metrics from the group of physical servers, which are evaluation indicators necessary for power efficiency calculation, and which include task execution amounts and power consumption of the group of physical servers, and stores them in a metrics collection DB in a memory unit; a metrics measurement interval determination unit that acquires the task execution amounts stored in the metrics collection DB, extracts changes in the task execution amounts at predetermined time intervals as frequency components, and determines the interval for metrics measurement using the extracted frequencies; a power efficiency calculation interval determination unit that determines, as the interval for power efficiency calculation, an interval that is equal to or greater than the determined interval for metrics measurement and equal to or less than a minimum executable interval that indicates the time from the start to the completion of power efficiency control; and a power efficiency calculation unit that calculates the power efficiency at the power efficiency calculation interval determined by the power efficiency calculation interval determination unit using the metrics collected by the metrics collection unit at the determined interval for metrics measurement. [Effects of the Invention]

[0016] According to the present invention, the intervals for measuring metrics and the intervals for calculating power efficiency can be appropriately determined in accordance with fluctuations in demand related to application usage. [Brief explanation of the drawings]

[0017] [Figure 1] 1 is a diagram showing the overall configuration of a power efficiency calculation system including a power efficiency calculation device according to an embodiment of the present invention. [Figure 2] 10A and 10B are diagrams illustrating an example in which frequency components are extracted at predetermined time intervals from data obtained by smoothing the task processing amount. [Figure 3] 10 is a flowchart showing the flow of processing executed by the power efficiency calculation device according to the present embodiment. [Figure 4] FIG. 2 is a hardware configuration diagram illustrating an example of a computer that realizes the functions of the power efficiency calculation device according to the present embodiment. [Figure 5] 10A and 10B are diagrams for explaining problems that arise when the frequency of metrics measurement and the frequency of power efficiency calculation are not set appropriately. DETAILED DESCRIPTION OF THE INVENTION

[0018] Next, an embodiment of the present invention (hereinafter referred to as "the present embodiment") will be described. FIG. 1 is a diagram showing the overall configuration of a power efficiency calculation system 1000 including a power efficiency calculation device 1 according to this embodiment. The power efficiency calculation system 1000 includes a group of physical servers 30 configured by a data center or the like, and a power efficiency calculation apparatus 1 connected to the group of physical servers 30 for communication.

[0019] The physical server group 30 is operated by, for example, building a virtualization infrastructure on a physical server, and one or more applications 3 are installed in a virtual machine (VM) or container on a virtual OS to execute processing. Each application 3 realizes a service by executing one or more tasks.

[0020] The power efficiency calculation device 1 collects metrics (such as "task execution amount" and "power consumption" described later) that are evaluation indices required for calculating DCeP (power efficiency) from the physical server group 30, and determines optimal intervals for measuring the metrics and for calculating the power efficiency that can capture changes in demand (corresponding to changes in the task execution amount of the application 3) processed by the application 3. Then, the power efficiency calculation device 1 performs metrics measurement and calculation of the power efficiency at the determined intervals for measuring the metrics and for calculating the power efficiency. As a result, the power efficiency calculation device 1 can suppress power waste and storage pressure caused by performing metrics collection processing and power efficiency calculation processing excessively frequently, and can also appropriately evaluate demand fluctuations of the target application.

[0021] Next, the power efficiency calculation device 1 will be described in detail. As shown in FIG. 1, the power efficiency calculation device 1 includes a control unit 10, an input / output unit 11, and a storage unit 12.

[0022] The input / output unit 11 inputs and outputs information to and from each server of the physical server group 30. The input / output unit 11 is composed of a communication interface that transmits and receives information via a communication line, and an input / output interface that inputs and outputs information to and from an input device such as a keyboard and an output device such as a monitor (not shown).

[0023] The storage unit 12 is configured by a hard disk, a flash memory, a RAM (Random Access Memory), or the like. The storage unit 12 temporarily stores programs for executing the functions of the control unit 10 and information necessary for the processing of the control unit 10. Furthermore, the storage unit 12 also stores metrics required for calculating DCeP (power efficiency) collected from each physical server in the physical server group 30, an OS (virtual OS), a VM / container, an application, etc. in a metrics collection database (DB: DataBase) 100. Furthermore, the storage unit 12 stores a minimum feasible interval (T c min ) information is stored. c min ) will be described in detail later.

[0024] The control unit 10 is responsible for all the processing executed by the power efficiency calculation device 1, and as shown in FIG. 1, is configured to include a metrics collection unit 101, a metrics measurement interval determination unit 102, a power efficiency calculation interval determination unit 103, a power efficiency calculation unit 104, and a power efficiency control unit 105.

[0025] The metrics collection unit 101 collects metrics (evaluation indexes for performance, etc.) required for power efficiency calculation from the physical server group 30 and stores them in the metrics collection DB 100. The metrics collector 101 collects metrics from the physical servers, OSs (virtual OSs), VMs / containers, applications, etc. that make up the physical server group 30 using existing metrics collection software (for example, Prometheus, etc.). For example, the metrics collector 101 collects information on task execution volume (number of requests processed by applications, amount of data transferred, etc.) and power consumption [W] as metrics from the physical server group 30 and stores it in the metrics collection DB 100. The metrics collection unit 101 collects metrics from the physical server group 30 in advance of calculating DCeP (power efficiency) (the "preparation stage" described below), and also collects metrics when calculating DCeP (power efficiency) (the "operation stage" described below).

[0026] The metrics measurement interval determination unit 102 extracts application demand fluctuations (changes in task execution volume) over a specified time interval as frequency components based on the task execution volume collected by the metrics collection unit 101, and determines the interval for metrics measurement using the extracted frequencies.

[0027] Specifically, the metrics measurement interval determination unit 102 acquires the actual measured values ​​(demand fluctuations for the application) of task processing volume (number of requests to the application, data transfer volume, etc.) for a specified period from the metrics collection DB 100, and performs smoothing (such as moving average) processing on the actual measured values ​​of task processing volume. The metrics measurement interval determination unit 102 then divides the smoothed data into predetermined time intervals and extracts frequency components from each of the intervals. The metrics measurement interval determination unit 102 then extracts the maximum frequency λ l Get (target frequency).

[0028] FIG. 2 shows an example of extracting frequency components for each window w of a predetermined time interval from data (the graph indicated by the dotted line p in FIG. 2) that has been smoothed (e.g., moved average) for task processing amounts shown in a histogram. The metrics measurement interval determination unit 102 determines the maximum (highest) frequency λ among the frequency components extracted at each predetermined time interval (window w). l Get (target frequency). Then, the metrics measurement interval determination unit 102 determines the target frequency λ based on the sampling theorem. l A frequency λ greater than twice the M Select the selected frequency λ M =1 / T M This determines the interval between metric measurements (T M ) is determined. Note that the metrics measurement interval determination unit 102 determines the frequency λ M When selecting the target frequency λ l A predetermined logic (for example, target frequency λ) is used to determine the frequency that is closest to double the target frequency λ. l A logic for setting the value to a predetermined multiple greater than twice the value (e.g., 2.5 times) is set in advance.

[0029] Returning to FIG. 1, the power efficiency calculation interval determination unit 103 determines the metric measurement interval (T M) based on the power efficiency calculation interval (T A ) is determined. The power efficiency calculation interval determination unit 103 determines the measurement interval (T M ) and the minimum feasible interval (T c min ) to calculate the power efficiency interval (T A ) is determined. Here, the minimum feasible interval (T c min ) means the time from when the power efficiency control is executed until the control is reflected after the execution of the control. c min ) means the time from the start to the completion of power efficiency control. For example, when an application on a container on a physical server is moved to a container on another physical server for power efficiency control, or when a physical server with a low utilization rate is put into a sleep state, the minimum executable interval (T c min ) 121 (see FIG. 1) is measured in advance before the operation stage processing is executed and stored in the storage unit 12.

[0030] The power efficiency calculation interval determination unit 103 determines the measurement interval (T M ) and T M ≦ T A Determine the value that Furthermore, when power efficiency control is performed, the power efficiency calculation interval determination unit 103 determines T A ≦ T c min Determine the value that T M ≦ T A ≦ T c min This determines the interval (T A ) can be used to determine upper and lower limits. Here, T c min ≦ T M If so, T M ≦ T AThe power efficiency calculation interval (T A ) is determined.

[0031] Power efficiency calculation interval (T A ) as the upper limit of the minimum feasible interval (T c min ) will be explained below. The purpose of power efficiency calculation is to improve power efficiency through some kind of control. To maximize power efficiency, it is desirable to detect inefficient states as soon as possible and intervene through control. If the interval between power efficiency calculations (T A ) is the minimum feasible interval (T c min ), the following two problems arise: (1) Although the next control is possible when a certain control is completed, it is not possible to determine whether or not the next control is necessary. (2) It is not possible to immediately perform a comparison between before and after some kind of control, which makes it difficult to distinguish between the effects of the control and other external factors (such as changes in the load). To avoid the above problem, the interval between power efficiency calculations (T A ) is the minimum feasible interval (T c min ) or less is desirable.

[0032] The power efficiency calculation unit 104 calculates DCeP (power efficiency) using metrics collected by the metrics collection unit 101 (for example, the number of processed requests, the amount of data transferred, power consumption, etc.). The power efficiency calculation unit 104 calculates the power efficiency of the system by calculating the power efficiency of the system when the metrics collection unit 101 calculates the power efficiency of the system at the interval (T M ) and acquires the metrics stored in the metrics collection DB 100. Then, the power efficiency calculation unit 104 calculates "Useful Work Produced" (amount of work result) using the above formula (2).

[0033] Furthermore, the power efficiency calculation unit 104 determines the power efficiency calculation interval (T A ), the total amount of power consumed by the physical server group 30 ("Total Energy Consumed to Perform that Work") is calculated. This total amount of power consumed by the physical server group 30 is calculated at the power efficiency calculation interval (T A ), the power consumption collected by the metrics collector 101 as metrics is summed up (the power efficiency calculation interval (T A ) is calculated by integrating it based on the amount of power consumed. Then, the power efficiency calculation unit 104 calculates the power efficiency by dividing the calculated "Useful Work Produced" (amount of work produced) by "Total Energy Consumed to Perform that Work" (amount of power consumption) using equation (1).

[0034] The power efficiency control unit 105 executes predetermined power efficiency control in accordance with the calculation result of the power efficiency calculated by the power efficiency calculation unit 104 . Here, the predetermined power efficiency control is, for example, a process in which, when power efficiency is reduced, an application on a server with low CPU utilization is moved to another server, and the server with low CPU utilization is put into sleep mode, etc. The content of this predetermined power efficiency control is set in advance according to the calculation result of power efficiency.

[0035] <Processing of power efficiency calculation device> Next, the flow of processing executed by the power efficiency calculation device 1 will be described. FIG. 3 is a flowchart showing the flow of processing executed by the power efficiency calculation device 1 according to this embodiment. The process executed by the power efficiency calculation device 1 is roughly divided into a preparatory stage and an operational stage. In the preparatory stage, metrics collected in advance from the physical server group 30 are used to calculate the measurement interval (T M ) and the interval for power efficiency calculation (T AThen, in the operation phase, the measurement interval (T M ), metrics are collected from the physical server group 30, and the collected metrics are used to calculate the power efficiency. A ) DCeP (power efficiency) is calculated for each Note that here, the minimum feasible interval (T c min ) 121 is assumed to be measured in advance and stored in the storage unit 12. As an application example, an example of monitoring DCeP (power efficiency) in the operating environment of a UPF (User Plane Function) application that transfers user data packets from a user terminal to a data network (such as the Internet) will be described.

[0036] First, in the advance preparation stage, the metrics collection unit 101 of the power efficiency calculation device 1 collects metrics (evaluation indexes for performance, etc.) required for power efficiency calculation from the physical server group 30 and stores them in the metrics collection DB 100 (step S10). The metrics collector 101 collects metrics using, for example, metrics collection software (Prometheus). The collected metrics include, for example, the amount of data transfer as the amount of task execution, and power consumption [W].

[0037] The metrics collection unit 101 may collect metrics by, for example, applying a test load to the application 3, or may collect metrics obtained by executing actual request processing, data transfer, etc. At this time, the metrics measurement interval by the metrics collector 101 is set as a provisional value, for example, one minute, which is the default value for metrics collection in Prometheus. The metrics collection unit 101 collects the task execution volume (data transfer volume) of the UPF application for a predetermined period, for example, one month.

[0038] Next, the metrics measurement interval determination unit 102 of the power efficiency calculation device 1 performs smoothing (moving average, etc.) processing (low-pass filter processing) on ​​the demand fluctuations (changes in task execution amount) of the application based on the task execution amount (data transfer amount, etc.) acquired by the metrics collection unit 101. Then, the power efficiency calculation device 1 divides the smoothed data into predetermined time intervals, extracts frequency components for each, and calculates the maximum frequency λ among the extracted frequencies. l (target frequency) is acquired (step S11).

[0039] Specifically, the metrics measurement interval determination unit 102 determines, for example, a daily fluctuation (a period focusing on one day) as the order of the control period, and performs smoothing processing by calculating a moving average of the data transfer volume over a predetermined time interval (for example, one hour).The metrics measurement interval determination unit 102 then divides the data into one-hour intervals with a fixed overlap period (for example, 30 minutes), and extracts frequency components from each.The metrics measurement interval determination unit 102 then calculates the maximum frequency λ of the extracted frequency components. l Get (target frequency).

[0040] Next, the metrics measurement interval determination unit 102 calculates the maximum frequency λ l (target frequency), based on the sampling theorem, the maximum frequency λ l Frequency λ greater than twice the target frequency M For frequency λ M =1 / T M This determines the measurement interval (T M That is, the metrics measurement interval determination unit 102 determines the maximum frequency λ l Frequency λ greater than twice the target frequency M The inverse of the metric measurement interval (T M ) (step S12).

[0041] In addition, if the task execution volume (data transfer volume) collected for one month shows a tendency for the task execution volume (data transfer volume) to increase from the first week to the fourth week of that month, the metrics measurement interval determination unit 102 may calculate a seven-day moving average, etc., and subtract that amount from the daily fluctuation data described above (high-pass filter processing).

[0042] Next, the power efficiency calculation interval determination unit 103 of the power efficiency calculation device 1 determines the metric measurement interval (T M ) and the minimum feasible interval (T c min ) 121 (see Figure 1) and the interval (T A ) is determined (step S13). Here, the power efficiency calculation interval determination unit 103 determines the measurement interval (T M ) or more, the interval for power efficiency calculation (T A ) to determine (T M ≦ T A ). The power efficiency calculation interval determination unit 103 determines the interval (T A ) to the minimum feasible interval (T c min )121 or less (T A ≦ T c min ) In other words, the power efficiency calculation interval determination unit 103 determines T M ≦ T A ≦ T c min The interval for calculating the power efficiency (T A ) is determined. The processes from steps S10 to S13 are carried out in the advance preparation stage.

[0043] Next, the processing in the operation stage will be described. In this operation stage, the metrics collection unit 101 determines the metrics measurement interval (T M ) to collect metrics (for example, data transfer volume, power consumption, etc.) (step S14).

[0044] Then, the power efficiency calculation unit 104 of the power efficiency calculation device 1 calculates the measurement interval (T M ) using the metrics (data transfer volume, power consumption, etc.) collected at the power efficiency calculation interval determination unit 103, the power efficiency calculation interval (T A ) to calculate DCeP (power efficiency) (step S15). The power efficiency calculation unit 104 calculates "Useful Work Produced" (amount of work result) using the above formula (2). The power efficiency calculation unit 104 also calculates the power efficiency calculation interval (T A ), the metrics collector 101 calculates "Total Energy Consumed to Perform that Work" (amount of power consumption) using the power consumption collected as metrics. The power efficiency calculator 104 then calculates the power efficiency by dividing "Useful Work Produced" (amount of work result) by "Total Energy Consumed to Perform that Work" (amount of power consumption) using the above formula (1).

[0045] Next, the power efficiency control unit 105 executes predetermined power efficiency control in accordance with the calculation result of the power efficiency calculated by the power efficiency calculation unit 104 (step S16). For example, the power efficiency control unit 105 executes processing such as sleep control for a server with a small amount of packet transfer.

[0046] In this way, the power efficiency calculation device 1 can appropriately determine the metrics measurement interval and the power efficiency calculation interval according to the demand fluctuations related to application usage. Therefore, the power efficiency calculation device 1 can suppress power waste and storage pressure caused by performing metrics collection processing and power efficiency calculation processing too frequently, and can appropriately evaluate the demand fluctuations of the target application.

[0047] <Hardware configuration> The power efficiency calculation device 1 according to this embodiment is realized by a computer 900 having a configuration as shown in FIG. 4, for example. 4 is a hardware configuration diagram showing an example of a computer 900 that realizes the functions of the power efficiency calculation device 1 according to this embodiment. The computer 900 includes a CPU (Central Processing Unit) 901, a ROM (Read Only Memory) 902, a RAM 903, an HDD (Hard Disk Drive) 904, an input / output I / F (Interface) 905, a communication I / F 906, and a media I / F 907.

[0048] The CPU 901 operates based on a program stored in the ROM 902 or the HDD 904, and performs control by the control unit 10 (FIG. 1). The ROM 902 stores a boot program executed by the CPU 901 when the computer 900 is started up, programs related to the hardware of the computer 900, and the like.

[0049] The CPU 901 controls an input device 910 such as a mouse or keyboard, and an output device 911 such as a display or printer, via an input / output I / F 905. The CPU 901 acquires data from the input device 910 via the input / output I / F 905, and outputs generated data to the output device 911.

[0050] The HDD 904 stores programs executed by the CPU 901 and data used by the programs. The communication I / F 906 receives data from other devices via a communication network (e.g., NW (Network) 920) and outputs the data to the CPU 901, and also transmits data generated by the CPU 901 to other devices via the communication network.

[0051] The media I / F 907 reads a program or data stored in the recording medium 912 and outputs it to the CPU 901 via the RAM 903. The CPU 901 loads a program related to a target process from the recording medium 912 onto the RAM 903 via the media I / F 907, and executes the loaded program. The recording medium 912 is an optical recording medium such as a DVD (Digital Versatile Disc) or a PD (Phase Change Rewritable Disc), a magneto-optical recording medium such as an MO (Magneto Optical Disk), a magnetic recording medium, a semiconductor memory, or the like.

[0052] For example, when a computer 900 functions as the power efficiency calculation device 1 according to this embodiment, a CPU 901 of the computer 900 executes a program loaded onto a RAM 903, thereby realizing the functions of the power efficiency calculation device 1. Furthermore, data in the RAM 903 is stored in an HDD 904. The CPU 901 reads and executes a program related to a target process from a recording medium 912. Alternatively, the CPU 901 may read a program related to a target process from another device via a communication network (NW 920).

[0053] <Effects> The effects of the power efficiency calculation device 1 and the like according to the present invention will be described below. The power efficiency calculation device according to the present invention is a power efficiency calculation device 1 that calculates the power efficiency resulting from the execution of an application 3 installed on a physical server group 30, and includes a metrics collection unit 101 that collects metrics, which are evaluation indexes necessary for power efficiency calculation and include task execution amounts and power consumption of the physical server group 30, from the physical server group 30 and stores the metrics in a metrics collection DB 100 in a storage unit 12; and a metrics collection unit 102 that acquires the task execution amounts stored in the metrics collection DB 100, extracts changes in the task execution amounts over a predetermined time interval as frequency components, and extracts the extracted metrics. The power efficiency control system is characterized by comprising: a metrics measurement interval determination unit 102 that determines the interval for metrics measurement using the determined frequency; a power efficiency calculation interval determination unit 103 that determines an interval for power efficiency calculation that is equal to or greater than the determined interval for metrics measurement and equal to or less than a minimum feasible interval 121 that indicates the time from the start to the completion of power efficiency control; and a power efficiency calculation unit 104 that calculates power efficiency at the interval for power efficiency calculation determined by the power efficiency calculation interval determination unit 103 using metrics collected by the metrics collection unit 101 at the determined interval for metrics measurement.

[0054] In this way, the power efficiency calculation device 1 can appropriately determine the metrics measurement interval and the power efficiency calculation interval according to the demand fluctuations related to application usage. Therefore, the power efficiency calculation device 1 can suppress power waste and storage pressure caused by performing metrics collection processing and power efficiency calculation processing too frequently, and can appropriately evaluate the demand fluctuations of the target application. Furthermore, by systematizing the procedure for determining the interval between metrics measurements and the procedure for determining the interval between power efficiency calculations, the power efficiency calculation device 1 is able to automatically determine the interval between metrics measurements and the interval between power efficiency calculations and calculate the power efficiency.

[0055] Furthermore, in the power efficiency calculation device 1, the metrics measurement interval determination unit 102 performs a smoothing process on the task execution amount acquired from the metrics collection DB 100, divides the smoothed task execution amount data into predetermined time intervals, extracts frequency components from each of the divided data, acquires the maximum frequency among the extracted frequencies, and determines the reciprocal of a frequency that is greater than twice the maximum frequency as the measurement interval of the metrics.

[0056] In this way, the power efficiency calculation device 1 extracts changes in task execution amount as frequency components and can determine the metric measurement interval based on the maximum frequency among the extracted frequencies. This allows the power efficiency calculation device 1 to determine a more appropriate metric measurement interval that matches the cycle of demand fluctuations. In other words, the power efficiency calculation device 1 can reproduce the cycle of demand fluctuations by measuring the metrics, and can determine a metric measurement interval that does not result in excessive measurement frequency.

[0057] The present invention is not limited to the above-described embodiments, and many modifications can be made by a person skilled in the art within the technical concept of the present invention. [Explanation of symbols]

[0058] 1. Power efficiency calculation device 3. Application 10 Control Unit 11 Input / output section 12 Storage section 30 physical servers 100 Metrics Collection DB 101 Metrics Collection Department 102 Metrics measurement interval determination unit 103 Power efficiency calculation interval determination unit 104 Power Efficiency Calculation Unit 105 Power efficiency control unit 121 Minimum Feasible Interval 1000 Power Efficiency Calculation System

Claims

1. A power efficiency calculation device that calculates power efficiency by executing an application installed on a group of physical servers, a metrics collection unit that collects metrics from the group of physical servers, which are evaluation indexes necessary for calculating power efficiency, and which include task execution amounts and power consumption of the group of physical servers, and stores the collected metrics in a metrics collection DB in a storage unit; a metrics measurement interval determination unit that acquires the task execution amount stored in the metrics collection DB, extracts a change in the task execution amount at a predetermined time interval as a frequency component, and determines an interval for metrics measurement using the extracted frequency; a power efficiency calculation interval determination unit that determines an interval for calculating power efficiency that is equal to or greater than the determined interval for measuring metrics and equal to or less than a minimum feasible interval indicating the time from the start to the completion of power efficiency control; and a power efficiency calculation unit that calculates the power efficiency at the power efficiency calculation interval determined by the power efficiency calculation interval determination unit using the metrics collected by the metrics collection unit at the determined metrics measurement interval; A power efficiency calculation device comprising:

2. The metrics measurement interval determination unit A smoothing process is performed on the task execution amount acquired from the metrics collection DB, the smoothed task execution amount data is divided into predetermined time intervals, frequency components are extracted from each of the divided data, the maximum frequency among the extracted frequencies is acquired, and the reciprocal of a frequency that is greater than twice the maximum frequency is determined as the measurement interval of the metrics.

2. The power efficiency calculation device according to claim 1,

3. A power efficiency calculation method for a power efficiency calculation device that calculates power efficiency by executing an application installed on a group of physical servers, comprising: The power efficiency calculation device collecting metrics from the group of physical servers, which are evaluation indices necessary for calculating power efficiency, and which include task execution amounts and power consumption of the group of physical servers, and storing the collected metrics in a metrics collection DB in a storage unit; acquiring the task execution amount stored in the metrics collection DB, extracting a change in the task execution amount at a predetermined time interval as a frequency component, and determining an interval for metrics measurement using the extracted frequency; determining an interval for calculating power efficiency that is equal to or greater than the determined interval for measuring metrics and equal to or less than a minimum feasible interval indicating the time from the start to the completion of power efficiency control; calculating the power efficiency at the determined power efficiency calculation interval using metrics collected at the determined metric measurement interval; A power efficiency calculation method comprising:

4. A power efficiency calculation system including: a group of physical servers on which applications that execute one or more tasks are installed; and a power efficiency calculation device that calculates power efficiency resulting from execution of the applications installed on the group of physical servers, The power efficiency calculation device a metrics collection unit that collects metrics from the group of physical servers, which are evaluation indexes necessary for calculating power efficiency, and which include task execution amounts and power consumption of the group of physical servers, and stores the collected metrics in a metrics collection DB in a storage unit; a metrics measurement interval determination unit that acquires the task execution amount stored in the metrics collection DB, extracts a change in the task execution amount at a predetermined time interval as a frequency component, and determines an interval for metrics measurement using the extracted frequency; a power efficiency calculation interval determination unit that determines an interval for calculating power efficiency that is equal to or greater than the determined interval for measuring metrics and equal to or less than a minimum feasible interval indicating the time from the start to the completion of power efficiency control; and a power efficiency calculation unit that calculates the power efficiency at the power efficiency calculation interval determined by the power efficiency calculation interval determination unit using the metrics collected by the metrics collection unit at the determined metrics measurement interval; A power efficiency calculation system comprising:

5. A program for causing a computer to function as the power efficiency calculation device according to claim 1 or 2.

Images