JP2005266845A - Computer use fee calculation device and computer processing division method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a computer use fee calculation device that can calculate, for every job of computer processing, a computer use fee reflecting an energy charge for energy consumption required in the computer processing. <P>SOLUTION: Computer processing being executed or having been executed by distribution over a plurality of computers 2 is divided between the computers 2. Computer processing in a management computer 4 executes a fee calculation process of multiplying a unit price per hour of energy consumed by each computer 2 and a processing time about each of a single or a plurality of processing steps divided in time series in the computer 2 to calculate an energy charge for each processing step, and summing up such energy charges to calculate an energy charge for the computer processing. The processing steps are preferably divided by a predetermined division process such that every single computer finishes the processing in a time zone of a fixed unit price per hour of energy. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a computer usage fee calculation apparatus and a computer arithmetic processing division method, and more specifically, considers the power cost in each computer of parallel arithmetic processing by a plurality of computers distributed on a network such as grid computing. The present invention relates to a computer usage fee calculation device and a computer arithmetic processing division method.

  Computer usage fees for providing computer processing services to external users are generally set separately for fixed basic fees and metered fees, and metered fees are charged for individual computer processing. Is done. In addition, the pay-per-use charge is generally calculated according to the unit price of time by the CPU time required for the computer calculation process and the used capacity of the storage device.

  In the case of parallel computing processing in which computer computing processing is divided and executed by multiple computers distributed on a network such as grid computing, the total usage fee for each computer is the computer usage fee for the computer computing processing. .

In the parallel processing as described above, depending on how the computer processing is divided into a plurality of processing units and which processing unit is assigned to which computer, the degree of congestion, processing speed, and CPU time of each computer Depending on the unit price, there is a difference in the speed (delivery date) and charge of computer calculation processing. Therefore, when priority is given to the delivery date, the computer arithmetic processing is divided and assigned so that as many arithmetic processing units as possible are not congested or assigned to a computer with a high processing speed. Also, if you want to keep the computer usage fee low, you can assign the divided processing unit to an inexpensive computer with a pay-as-you-go fee, and optimize the computer processing including the communication cost for transferring the processing unit at that time to each computer A process division / allocation method has been proposed (for example, see Patent Document 1 below).
Special table 2001-503170 gazette

  The basic fee or metered fee for using the computer is determined by allocating the basic fee or metered fee, for example, referring to past performance, so that some or all of the cost required to operate the computer is borne by the user. . Therefore, as a cost required for the operation of the computer, a large rate of electricity for operating the computer accounts for a large proportion in a large computer or a supercomputer.

  By the way, in the case of the above parallel operation processing, since a plurality of computers used for the operation processing are distributed on the network, when the network is spread all over the world such as the Internet, the plurality of computers are also Will be distributed throughout the world. In addition, a plurality of arithmetic processing units processed at the same time are processed in the daytime in one computer and processed in the nighttime in another computer. Therefore, the power charge in the execution time zone of each arithmetic processing unit varies depending on the region and time zone for each computer. Further, even when the computer arithmetic processing is not divided into parallel arithmetic processing and processed by one computer, the power charge required for the arithmetic processing varies depending on the execution time zone.

  However, in the recent global trend of electricity liberalization, the electricity charge varies depending on the region and time, whereas the metered charge for each computer operation processing is the power consumption required for the operation processing. The current situation is that the charging is not fair reflecting the corresponding power charge (or energy charge).

  The present invention has been made in view of the above-mentioned problems, and a first object of the present invention is to solve the above-mentioned problems and to calculate a computer usage fee that reflects an energy fee for the amount of energy used for computer calculation processing. An object of the present invention is to provide a computer usage fee calculation device that can calculate for each computer calculation process. In addition, the second object of the present invention is to provide a computer calculation process that can optimize a computer usage fee when a computer usage fee that reflects an energy fee for the energy usage required for the computer calculation processing is charged. An object of the present invention is to provide a computer arithmetic processing division method for inputting to a plurality of computers.

  The computer usage charge calculation apparatus according to the present invention for achieving the first object is consumed by a computer separately for each of a plurality of operation processing pieces that are divided into time-series computer operation processes being executed or executed. A charge calculation process for calculating the energy charge of the computer calculation process by multiplying the energy unit price of energy and the calculation processing time and calculating the energy charge of each calculation processing piece is executed by the management computer calculation process. This is a first feature.

  Further, the computer usage fee calculation apparatus divides computer calculation processing being executed or executed in a distributed manner among a plurality of computers by the computer, and the calculation processing is divided into one or a plurality of time series for each computer. Charge calculation for calculating the energy charge of the computer calculation process by calculating and summing the energy charge of each calculation processing piece by multiplying the unit price of energy consumed by the computer and the calculation processing time separately for each piece A second feature is that the process is executed by a management computer calculation process.

  Further, in addition to the first or second feature, the computer usage fee calculation apparatus is configured such that the arithmetic processing of each arithmetic processing piece falls within a certain time zone in which the unit price of energy for the computer executing the arithmetic processing piece is constant. As described above, a third feature is that the processing for classifying the computer arithmetic processing into the arithmetic processing pieces is executed by the management computer arithmetic processing.

  According to the computer usage fee calculation apparatus of the first feature described above, when the unit price of energy required for the operation of the computer varies depending on the time zone, for example, a computer operation process by one computer is performed in a plurality of operations along a time series. By dividing into processing pieces, calculating the energy charge of each processing piece by multiplying the unit price of energy in the time zone to which each processing piece belongs and the processing time, and summing the energy charges of each processing piece, It is possible to calculate the energy charge for the entire computer processing. As a result, based on this energy fee, the computer usage fee reflecting the energy cost of the energy consumed by the computer by the computer arithmetic processing can be charged to the user. Therefore, for example, a user who performs a computer calculation process using an inexpensive late-night charge at night can enjoy the merit of using the inexpensive late-night charge.

  Furthermore, according to the computer usage fee calculation apparatus of the second feature, when the energy unit cost required for the computer operation differs depending on the location of the computer, or when the energy unit cost required for the computer operation varies depending on the time zone, a plurality of The computer arithmetic processing that is being executed or executed in a distributed manner is also divided into one or a plurality of arithmetic processing pieces along the time series for each computer, and the unit time of energy in the time zone to which each arithmetic processing piece belongs The energy charge of each calculation processing piece is calculated by multiplying the calculation processing time and the energy charge of each calculation processing piece is totaled, whereby the energy charge of the entire computer calculation processing can be calculated. As a result, for computer computation processing that is being executed or executed in a distributed manner among multiple computers, the computer usage fee that reflects the energy cost of the energy consumed by the computer through the computer computation processing is also calculated based on this energy fee. Can be charged.

  Furthermore, according to the computer usage fee calculation device of the third feature, each calculation processing piece is defined so that the unit price of energy in the time zone to which each calculation processing piece belongs is constant. It is possible to calculate the energy charge of a piece by simple multiplication.

  Here, the computer calculation process for management means a computer calculation process executed by the computer usage charge calculation device, and is different from the computer calculation process for which the usage charge is calculated. The energy unit price is an energy fee per unit energy usage of the computer, and the energy hourly unit means an energy fee per unit arithmetic processing time of each arithmetic processing piece. Therefore, if the energy unit price is high, the hourly unit price of energy increases in proportion. Further, the divided arithmetic processing pieces are units for calculating energy charges, and do not necessarily match the arithmetic processing units.

  In order to achieve the second object, the computer arithmetic processing division method according to the present invention includes a plurality of parallel or tandem arithmetic processing units or parallel and tandem arithmetic processing in order to input the computer arithmetic processing to a plurality of computers. A computer arithmetic processing division method that divides into units, a division step of dividing the computer arithmetic processing into a plurality of arithmetic processing units according to the respective computers and arithmetic processing time zones, and a predicted arithmetic processing amount for each arithmetic processing unit An arithmetic processing amount prediction step for calculating the calculation processing amount, and multiplying the sum of the unit time of energy consumed by the computer and the unit time of CPU time for each arithmetic processing unit by the prediction arithmetic processing time derived from the prediction arithmetic processing amount. The calculation processing fee of the calculation processing unit is calculated, and the calculation processing fee is calculated by summing up the calculation processing fee of each calculation processing unit until at least the next synchronization timing. A calculation process fee for calculating a calculation process charge of the calculation process, and a calculation of the computer calculation process predicted by the charge prediction process by associating the plurality of computers with each of the calculation processing units divided by the division process. An optimization step for determining the assignment of the computer to the arithmetic processing unit that minimizes the processing fee; and an input step for introducing the corresponding arithmetic processing unit to each of the computers assigned in the optimization step; The first feature is that the above is executed by computer processing for management.

  According to the computer arithmetic processing dividing method of the first feature, when the computer usage fee reflecting the energy cost of the energy consumed by the computer by the computer arithmetic processing is charged to the user, it is divided in the dividing step. The calculation fee that reflects the energy cost for the calculated computer calculation process is predicted in the charge prediction process, and in the optimization process, the computer allocation to the calculation processing unit with the minimum calculation process charge is determined. In the computer arithmetic processing processed by the arithmetic processing unit input to the computer, the arithmetic processing fee is optimized.

  Further, in addition to the first feature, the computer arithmetic processing division method re-executes the optimization step for each synchronization timing of each arithmetic processing unit divided by the division step, The second feature is that the corresponding arithmetic processing unit is input to each of the computers allocated in the re-executed optimization step.

  According to the computer arithmetic processing division method of the second feature, since the actual arithmetic processing time and elapsed time of each process change depending on the load status of each computer at the input destination, the computer optimized first Is not necessarily in the optimized state at the next synchronization timing, so the optimization process is executed at each synchronization timing, so that all processes are finally processed The total fee will be optimized according to the calculation progress.

  Further, in addition to the first or second feature, the computer arithmetic processing division method is a correspondence between a communication fee required for the input step, and each computer assigned in the optimization step and the arithmetic processing unit. A second fee calculation step of calculating a second calculation processing fee determined by the sum of the calculation processing fees of the computer calculation processing based on a relationship; and the computer calculation processing for each of the calculation processing units divided in the division step A second calculation processing time prediction step for calculating a second prediction calculation processing time for each calculation processing unit, and for each calculation processing unit divided in the division step And multiplying the sum of the unit price of energy consumed by the computer in which the computer calculation process exists and the unit price of CPU time by the second prediction calculation processing time, and calculating the calculation processing fee for each calculation processing unit. A third fee calculation step of calculating a third calculation processing fee of the computer calculation processing by summing up the calculation processing fees of the calculation processing units until at least the next synchronization timing, A third feature is that the process is further executed by computer calculation processing, and the charging step is executed when the second calculation processing fee is lower than the third calculation processing fee.

  According to the computer arithmetic processing division method of the third feature, the parallel arithmetic processing can be performed at a low cost in consideration of the communication fee required to input each arithmetic processing unit to each computer assigned in the optimization step. Since the parallel operation process is started when it can be executed, if the cost is prioritized over the speed of the operation process, if the parallel operation process cannot be executed at the minimum cost, the execution is suspended. As a result, it is possible to realize cost optimization taking into consideration the communication charge in parallel arithmetic processing.

  Furthermore, in addition to the first, second, or third feature, the computer arithmetic processing division method receives information about the arithmetic processing time of the arithmetic processing unit executed by each computer from the respective computers, and the arithmetic processing unit The calculation processing fee for each calculation processing unit is calculated by multiplying the sum of the unit time of energy consumed by the computer and the unit time of CPU time by the calculation processing time, and the calculation processing fee for each calculation processing unit is all A fourth feature is that the charge calculation step of calculating the calculation processing fee of the computer calculation processing by adding the values is further executed by the management computer calculation processing.

  According to the computer arithmetic processing dividing method of the fourth feature, when the computer usage fee reflecting the energy cost of the energy consumed by the computer by the computer arithmetic processing is charged to the user, it is divided in the dividing step. An arithmetic processing fee that reflects the energy cost for the computer arithmetic processing that is actually arithmetically processed is calculated in the fee calculating step. As a result, based on this energy fee, the computer usage fee reflecting the energy cost of the energy consumed by the computer by the computer arithmetic processing can be charged to the user.

  In order to achieve the first object, a computer usage charge calculation program according to the present invention provides a computer usage charge calculation apparatus having the first or second feature, wherein the classification process and the charge calculation process are performed by a computer. Each has a computer program that runs on it.

  According to the computer usage fee calculation program, the computer usage fee calculation device according to the present invention that exhibits the above-described effects can be realized on the computer by executing each computer program on the computer.

  In order to achieve the second object, a computer arithmetic processing division program according to the present invention performs processing of each step in the computer arithmetic processing division method having any one of the fourth to seventh features on a computer. Each has a computer program to be executed.

  According to the computer arithmetic processing division program, by executing each computer program on a computer, the computer arithmetic processing division method according to the present invention that exhibits the effects of the above features can be implemented on the computer.

  DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of a computer usage charge calculation apparatus (hereinafter referred to as “the present invention apparatus”) and a computer arithmetic processing division method (hereinafter referred to as “the present invention method”) according to the present invention will be described with reference to the drawings. To do.

  In this embodiment, as shown in FIG. 1, a single computer calculation process (hereinafter referred to as “job” for convenience) composed of a calculation program and data is a plurality of calculation processing units (hereinafter referred to as “job”). For convenience, it is divided into “processes”) and executed in parallel on a plurality of computers 2 (computers) distributed and connected to each other on a computer network 3 such as the Internet as schematically shown in FIG. This will be described assuming grid computing. Here, the parallel process means that each process is executed in parallel in a plurality of computers, and the cascaded process means that each process is executed before and after the same computer or between different computers. Means the case.

<First Embodiment>
The device 1 of the present invention is realized by executing a computer usage fee calculation program that constitutes a part of a grid computing management program executed on a management computer 4 that manages grid computing.

  The computer usage charge calculation program includes a division processing program step for dividing a series of processes assigned to each computer 2 into a plurality of operation processing pieces arranged in chronological order, and power consumed by the computer 2 for each operation processing piece (energy ) By calculating the energy charge of each operation processing piece by multiplying the time unit price and the operation processing time, and calculating the energy charge of the job.

  The present invention device 1 is a device that calculates an energy fee according to the power consumption of each computer for a job that has already been processed or a job that is already being submitted to a plurality of computers, and according to the amount of calculation processing of the job In addition, a part or all of the usage fee can be charged by the energy fee. In the first embodiment, cost optimization processing in the first embodiment to be described later is not considered. In the following description, it is assumed that each of a series of processes executed separately by each computer is processed in the same time zone in which the unit price of energy is the same, Assume that the processes match. The operation and processing procedure of the device 1 of the present invention will be described below with reference to the flowchart of FIG.

  First, the grid computing management program which is the main body of the device 1 of the present invention divides a job into a plurality of parallel and tandem processes as illustrated in FIG. 1 (step # 1). Here, the first, 1, 2, 3,... In the notation of each process in FIG. 1 as the process 2A, the process 5Ca, etc. represents the calculation processing order along the time series of each process. , B, C,... Distinguish between computers. When there is no computer, processing by one computer is shown, and the third a and b (which may not be present) distinguish parallel processing on the same computer. In the example of FIG. 1, the processes 3A, 3B, 3C and the processes 4A, 4B, 4C, the processes 5A, 5B, 5Ca, 5Cb and the processes 6A, 6B, and the processes 6A, 6B, and the process 7 A synchronization point has occurred. At the synchronization point, each process in the next stage mutually uses the calculation processing result of each process in the previous stage, so it is necessary to wait for the end of the calculation process in each process in the previous stage.

  Next, the context of each divided process is extracted and stored in a predetermined database (step # 2). Subsequently, the calculation amount (CPU load) of each process is calculated and stored in the database (step # 3).

  Next, for each process input destination (computer) candidate, the operation status (whether it is operating, considering the case of operation stoppage due to regular maintenance, etc.), availability (congestion level), and time zone The energy time unit price E0 (t) and the CPU time unit price C0 (t) are inquired, and the results are tabulated and stored (step # 4). “T” indicates time zone distinction. FIG. 4 schematically shows information to be acquired. This inquiry may be made in advance. In the present embodiment, the metered charge of the computer usage charge is proportional to the CPU time, and the energy charge reflecting the energy unit price at the time of the calculation process is proportional to the CPU time, and the energy unit price at the time of the calculation process. Is assumed to be configured with a CPU charge unrelated to the.

  Next, the assignment of the input destination (computer) for executing each process is determined (step # 5). In this assignment, optimization processing differs depending on whether processing speed is given priority or calculation cost is given priority. For example, when priority is given to processing speed, processes with a high CPU load are assigned in order from a computer with a low congestion degree.

  Next, each process is input to each input destination assigned in step # 5 (step # 6). Each input process is processed by the input destination computer. Each computer records the CPU time and the elapsed time (start time and end time) required for the arithmetic processing of each process, and the process in each computer is recorded. Information is transmitted to the device 1 of the present invention every time the process ends or when all processes are completed. The device 1 of the present invention receives the information and executes a charge calculation process (step # 7).

  For each process, the charge calculation processing is performed by multiplying the CPU time T (i) of each process, the computer in which it is processed, the energy time unit price E0 (t) of the time zone, and the CPU time unit price C0 (t). The pay-per-use fee P1 (i) for each process given by 1 is calculated. Here, “i” represents a distinction between a process or an operation processing piece. In Equation 1, E1 (i) represents the energy charge in each process.

(Equation 1)
P1 (i) = E1 (i) + C1 (i)
E1 (i) = T (i) × E0 (t)
C1 (i) = T (i) × C0 (t)

  When the metered fee P1 (i) of each process is summed for all the processes, as shown in Equation 2, a computer usage fee (metered fee) P2 for the entire job can be calculated.

(Equation 2)
P2 = ΣP1 (i)

  In the above description, it is assumed that the arithmetic processing piece generated by the classification processing matches the process, but an example of processing when the energy hour unit price E0 (t) changes during the arithmetic processing of some processes. Will be explained.

  As a first example, it is determined whether the elapsed time (start time and end time) of each process includes a change point in the energy hour unit price E0 (t). The process is divided into two arithmetic processing pieces according to the elapsed time before and after, and the CPU time of each arithmetic processing piece is apportioned in proportion to each elapsed time ratio. Accordingly, the divided processing pieces are handled in the same manner as the process, and the computer usage fee (usage fee) P2 for the entire job is calculated using the formulas (1) and (2).

  As a second example, when each computer accumulates the CPU time of each process, the CPU time at the changing point of the energy time unit price E0 (t) of each process is recorded, and the CPU time of each process is determined as the elapsed time. When transmitting together with the CPU time at the change point. Therefore, it is determined whether the elapsed time (start and end times) of each process includes a change point in the energy hour unit price E0 (t). If the change time is included, the elapsed time before and after the change point. Then, the process is divided into two arithmetic processing pieces, and the CPU time of each arithmetic processing piece is the CPU time at the received change point as the CPU time of the first half arithmetic processing piece, and the change point from the CPU time of the entire process. The CPU time obtained by subtracting the CPU time in is used as the CPU time of the latter half of the arithmetic processing piece. Then, the divided processing pieces are handled in the same manner as the process, and the computer usage fee (payment fee) P2 for the entire job is calculated using the formulas (1) and (2).

Second Embodiment
Next, the computer arithmetic processing division method (the present invention method) according to the present invention will be described. The method of the present invention is realized by executing a computer arithmetic processing division program constituting a part of the grid computing management program on the management computer (see FIG. 2).

  The computer arithmetic processing division program is a program step for executing a division process for dividing a job into a plurality of processes according to each computer 2 and an arithmetic processing time zone, and an arithmetic operation for calculating a predicted arithmetic processing amount (predicted CPU load) for each process. Each of the program steps for executing the processing amount prediction step, and the sum of the unit time of energy consumed by the computer for each process and the unit time for CPU time multiplied by the predicted CPU time obtained by dividing the predicted CPU load by the processing speed of the computer. A program step for executing a charge prediction process for calculating a calculation processing fee for a computer calculation process by calculating a calculation processing fee for a process and totaling the calculation processing charges for each process until at least the next synchronization timing, and a dividing step Make multiple computers correspond to each divided processing unit and make predictions using the charge prediction process The program step that executes the optimization process that determines the allocation of the computer to each process, and the process that corresponds to each computer allocated in the optimization process, which minimizes the calculation processing fee of the computer calculation process Receives information about the program step for executing the process and the CPU time of the process executed by each computer from each computer, and multiplies the actual CPU time by the sum of the unit time of energy consumed by the computer and the unit time of CPU time for each process. And a program step for executing a charge calculation step for calculating a calculation processing fee for each process by calculating a calculation processing fee for each process and summing all the calculation processing fees for each process.

  In the method of the present invention, when a job is divided into a plurality of processes and each process is allocated to a plurality of computers for parallel operation processing, a pay-per-use fee for the calculation processing considering the energy fee according to the power consumption of each computer is optimized. In this way, each process is associated with each computer under optimum conditions. In the following description, it is assumed that each of a series of processes executed separately on each computer is processed within the same time zone in which the unit price of energy is the same. The processing procedure of the method of the present invention will be described below with reference to the flowchart of FIG.

  First, a computer arithmetic processing division program (corresponding to the grid computing management program of the first embodiment) for executing the method of the present invention divides a job into a plurality of parallel and tandem processes as illustrated in FIG. # 1).

  Next, the context of each divided process is extracted and stored in a predetermined database (step # 2). Subsequently, the calculation amount (CPU load) K ′ (i) of each process is predicted and stored in the database (step # 3). “I” represents the distinction between processes.

  Next, for each process input destination (computer) candidate, the operation status (whether it is operating, considering the case of operation stoppage due to regular maintenance, etc.), availability (congestion level), and time zone The energy time unit price E0 (t) and the CPU time unit price C0 (t) are inquired, and the results are tabulated and stored (step # 4). “T” indicates time zone distinction. FIG. 4 schematically shows information to be acquired. This inquiry may be made in advance. In the present embodiment, the metered charge of the computer usage charge is proportional to the CPU time, and the energy charge reflecting the energy unit price at the time of the calculation process is proportional to the CPU time, and the energy unit price at the time of the calculation process. Is assumed to be configured with a CPU charge unrelated to the.

  Next, the assignment of the input destination (computer) for executing each process is determined (step # 5). In this embodiment, in this allocation, an optimization process giving priority to the calculation cost is executed. For example, a process with a large CPU load is assigned in order from a computer having a low total of the energy time unit cost E0 (t) and the CPU time unit cost C0 (t) in the input time zone. Here, the process to be assigned is a process to be processed before the next synchronization timing. As shown in FIG. 6, when process allocation is performed in order starting from the computer having the lowest total energy time unit cost E0 (t) and CPU time unit cost C0 (t) and ignoring the congestion degree, Since the computation processing elapsed time in the computer may be abnormally long compared to other computers, considering the predicted CPU time T ′ (i) and the degree of congestion, a plurality of processes and computer processes are performed according to a predetermined combination algorithm. A correspondence relationship is extracted (step # 51). Next, based on each correspondence relationship, the CPU time T ′ (i) of each process is calculated by dividing the calculation amount K ′ (i) of each process predicted in step # 3 by the processing speed of the corresponding computer. (Step # 52). The calculated CPU time T ′ (i) is a predicted value, and a considerable error may occur depending on the contents of the job. Therefore, the predicted value is only a guide. Then, a charge prediction step is executed for each extracted correspondence (step # 53). Then, from the extracted plurality of correspondence relationships, the correspondence relationship having the minimum total of the processing charges P1 '(i) of the processes processed until the next synchronization timing is extracted (step # 54).

  In the charge prediction step, first, for each process, the predicted CPU time T ′ (i) of each process, the computer to be allocated, the energy time unit price E0 (t) of the time zone, and the CPU time unit price C0 (t) are multiplied. , The metered fee P1 ′ (i) for each process given by Equation 3 is calculated, and the computer usage fee (usage fee) P2 ′ for the entire process processed by the calculation formula given by Equation 4 until the next synchronization timing is calculated. Is calculated.

(Equation 3)
P1 ′ (i) = E1 ′ (i) + C1 ′ (i)
E1 ′ (i) = T ′ (i) × E0 (t)
C1 ′ (i) = T ′ (i) × C0 (t)

(Equation 4)
P2 ′ = ΣP1 ′ (i)
However, i covers all processes processed before the next synchronization timing.

  Next, each process up to the next synchronization timing is input to each input destination assigned in step # 5 (step # 6). Each input process is processed by the input computer, but at the next synchronization timing, the execution of each input process is completed, so that it waits for the input of the next process. Become. Accordingly, the arrival of the next synchronization timing is determined (step # 7). At the synchronization timing, the processing of step # 5 is repeated, and the next processes are input to the newly assigned computer (step # 6). .

  If it is determined in step # 7 that the synchronization timing has arrived, the calculation process of step # 8 is entered.

  Each process submitted at the start of the job and each synchronization timing is processed by the computer to which the job is submitted, and each computer is responsible for the CPU time and the elapsed time (start and end) required for each process. Time) is recorded, and each time the process in each computer is completed, the information is transmitted to the management computer that executes the method of the present invention at the time of synchronization timing or when all processes are completed. The management computer receives the information and executes a charge calculation step (step # 8).

  In the charge calculation process, each process is multiplied by the actual CPU time T (i) of each process, the computer in which it is processed, the energy time unit price E0 (t) of the time zone, and the CPU time unit price C0 (t). , A metered fee P1 (i) for each process given by Equation 1 is calculated, and a computer usage fee (usage fee) P2 for the entire job is calculated using the formula given by Equation 2. Note that the charge calculation process is the same as the charge calculation process of the first embodiment, and therefore a duplicate description is omitted. In the above description, it is assumed that each of a series of processes executed separately by each computer is processed in the same time zone in which the unit price of energy is the same. The process when the energy hour unit price E0 (t) changes can be similarly processed in the manner described in the first embodiment.

<Third Embodiment>
Next, another embodiment of the method of the present invention will be described. In addition to the configuration of the second embodiment, the computer arithmetic processing division program for realizing the method of the present invention in the third embodiment further includes a communication fee required for the input process, each computer and each process allocated in the optimization process. And a program step for executing a second fee calculation step for calculating a second calculation processing fee determined by the sum of the calculation processing fee of the job based on the correspondence relationship with A program step for executing a second calculation processing time prediction step for calculating a second prediction calculation processing time for each process, assuming that the calculation processing is performed only by an existing computer, and a job for each process divided in the division step Multiply the sum of the hourly unit price of energy consumed by the computer where the computer exists and the unit price of CPU time by the predicted computation processing time of the computer, and calculate each process. A program step for executing a third fee calculation step for calculating a third fee for calculating a job by calculating a fee for calculating a third fee for a job by calculating a fee for calculating a third fee for a job by calculating a fee for calculating a third fee Composed.

  The difference between the third embodiment and the second embodiment of the method of the present invention is that, after the computer assignment process (optimization process) to each process of step # 5 in the flowchart shown in FIG. Before, as shown in FIG. 7, it is a point which implements an insertion determination process.

  The input determination process includes a communication charge D required for the input process and an operation processing charge P2 ′ for all processes until the next synchronization timing based on the correspondence between each computer and each process allocated in the optimization process. A second fee calculation step of calculating a second calculation processing fee P3 (= D + P2 ′) determined in total is executed (step # 55). Note that the calculation processing fee P2 'may be the one calculated in the fee prediction step.

  Next, a second prediction CPU for each process when it is assumed that the process divided in the division step is processed only by the computer in which the job exists without transferring the process to the other computer 2 via the computer network 3. A second calculation processing time prediction step for calculating time T ″ (i) is executed (step # 56).

  Next, the sum of the time unit cost E0 (t) and the CPU unit time unit C0 (t) of energy consumed by the computer where the job exists is multiplied by the second predicted CPU time T ″ (i) in the computer, The calculation processing fee P3 (i) is calculated by the calculation formula shown in Equation 5, and the calculation processing fees P3 (i) of all the processes up to the next synchronization timing are summed as shown in Equation 6, and each process concerned is calculated. A third fee calculation step for calculating the entire processing fee P4 is executed (step # 57).

(Equation 5)
P3 (i) = E3 (i) + C3 (i)
E3 (i) = T ″ (i) × E0 (t)
C3 (i) = T ″ (i) × C0 (t)

(Equation 6)
P4 = ΣP3 (i)
However, i covers all processes processed until the next synchronization timing.

  Next, the calculation processing fee P3 (= D + P2 ′) including the communication fee calculated in the second fee calculation step is compared with the calculation processing fee P4 calculated in the third fee calculation step (step # 58). When the calculation processing fee P3 is equal to or less than the calculation processing fee P4, the process proceeds to the input process (step # 6). When the calculation processing fee P3 is higher than the calculation processing fee P4, the input timing is changed (step # 59), the process returns to step # 5, and the optimization process and the input determination process are executed again. For example, if the calculation processing fee P3 does not fall below the calculation processing fee P4 even after being re-executed a predetermined number of times (step # 60), the process is not transferred to another computer 2, that is, the process is stopped and the job is stopped. The calculation processing is performed only by the computer in which the user exists, or a job interruption warning is issued to the user. Since the processing after the job is submitted is the same as that in the second embodiment, a duplicate description is omitted.

  Hereinafter, another embodiment will be described.

  <1> In the above embodiment, as shown in Equations 1 and 2, the job-based fee is an energy fee proportional to the amount of energy (electric power) consumed by each computer, and the CPU time regardless of the amount of energy. However, the usage fee may be composed only of the energy fee. That is, the present invention can be applied to a fee system that collects a cost other than the energy cost required for the operation of the computer with a fixed basic fee.

  <2> In the third embodiment, an example in which the calculation processing fee P3 and the calculation processing fee P4 are compared in the insertion determination step has been shown. For example, the proportion of the communication fee D in the calculation processing fee P3 is When it is less than or equal to the predetermined percentage, the process may be shifted to the charging process.

  <3> In the second and third embodiments, the prediction accuracy of the CPU time T ′ (i) may be a problem depending on the contents of the job. Therefore, the predicted CPU time T ′ (i) is temporarily exceeded. It is also preferable to issue a warning to the user in advance when the execution of the arithmetic processing of some processes is prolonged.

  <4> In the second and third embodiments described above, an upper limit value for the elapsed time for the arithmetic processing of the job is determined, and in the optimization step, the degree of congestion and load factor of the computer executed by each process and the predicted CPU time Predict the elapsed time of each process, predict the total elapsed time of the entire job based on the context of each process stored in the database, and in addition to cost optimization, ensure that the total elapsed time does not exceed the upper limit It is also a preferred embodiment to perform optimization.

  <5> In the first embodiment, the computer usage charge calculation apparatus according to the present invention has been described for the case where it is applied to the calculation of the energy charge assuming grid computing. However, the calculation of the energy charge with one computer is described. It is also applicable to.

The figure explaining a mode that a computer arithmetic processing is divided | segmented into the several arithmetic processing unit parallel and cascaded The figure which shows the interconnection relation of the computer which is used for grid computing The flowchart which shows the operation | movement procedure in one Embodiment of the computer usage fee calculation apparatus which concerns on this invention. The figure which shows typically an example of the empty condition (congestion degree) of a computer, the energy time unit price according to a time slot | zone, and CPU time unit price The flowchart which shows the process sequence in one Embodiment of the computer arithmetic processing division | segmentation method concerning this invention. The flowchart which shows the process sequence in the optimization process of the computer arithmetic processing division | segmentation method based on this invention. The flowchart which shows the process sequence of the input determination process in other embodiment of the computer arithmetic processing division | segmentation method concerning this invention.

Explanation of symbols

1: Computer usage fee calculation apparatus according to the present invention 2: Computer 3: Computer network 4: Management computer

Claims (9)

  Calculate the energy charge for each processing piece by multiplying the time unit of energy consumed by the computer and the processing time for each of the processing pieces that are divided into time series of the computer processing that is being executed or executed. Then, a computer usage charge calculation apparatus that executes a charge calculation process for calculating the energy charge of the computer calculation process by performing a summation by a management computer calculation process.   The computer computation processing being executed or executed by being distributed to a plurality of computers is classified by the computer, and the computer consumes the computation processing pieces divided into one or more in time series for each computer. The charge calculation processing for calculating the energy charge of the computer calculation processing by calculating the energy charge of each calculation processing piece by multiplying the unit time of energy and the calculation processing time and summing them is performed by the computer calculation processing for management. A computer usage charge calculation device characterized by being executed.   The management process is a classification process that divides the computer arithmetic processing into the arithmetic processing pieces so that the arithmetic unit of the arithmetic processing pieces is within a fixed time period. The computer usage fee calculation apparatus according to claim 1, wherein the computer usage charge calculation device is executed by a computer calculation process for a computer. A computer arithmetic processing division method that divides a plurality of parallel or tandem arithmetic processing units or parallel and tandem arithmetic processing units in order to input computer arithmetic processing to a plurality of computers,
A division step of dividing the computer arithmetic processing into a plurality of arithmetic processing units according to the respective computers and arithmetic processing time zones;
An arithmetic processing amount prediction step of calculating a prediction arithmetic processing amount for each arithmetic processing unit;
An arithmetic processing fee for each arithmetic processing unit is calculated by multiplying the sum of the hourly unit cost of energy consumed by the computer and the unit cost of CPU time for each arithmetic processing unit by the predicted arithmetic processing time derived from the predicted arithmetic processing amount. A fee prediction step of predicting the calculation processing fee of the computer calculation processing by summing up the calculation processing fees of each calculation processing unit until at least the next synchronization timing;
The computer for the arithmetic processing unit, wherein the plurality of computers correspond to the arithmetic processing units divided by the dividing step, respectively, and the arithmetic processing fee of the computer arithmetic processing predicted by the fee prediction step is minimized. An optimization process that determines the allocation of
An input step of inputting the corresponding arithmetic processing unit to each of the computers allocated in the optimization step,
Is executed by a management computer calculation process for management.   The optimization step is re-executed at each synchronization timing of each arithmetic processing unit divided by the division step, and each of the computers assigned in the re-execution optimization step is re-executed in the input step, respectively. 5. The computer arithmetic processing division method according to claim 4, wherein the corresponding arithmetic processing unit is input. A second arithmetic processing fee determined by the sum of the communication fee required for the input step and the arithmetic processing fee of the computer arithmetic processing based on the correspondence between each computer assigned in the optimization step and the arithmetic processing unit A second fee calculating step for calculating
A second calculation for calculating a second predicted calculation processing time for each calculation processing unit when it is assumed that calculation processing is performed only for the computer in which the calculation processing exists for each calculation processing unit divided in the dividing step. Processing time prediction step;
For each of the arithmetic processing units divided in the dividing step, each arithmetic processing unit is obtained by multiplying the sum of the unit price of energy consumed by the computer in which the computer arithmetic processing exists and the unit price of CPU time by the second predicted arithmetic processing time. A third fee calculation step of calculating a third calculation processing fee of the computer calculation processing by calculating the calculation processing fee of the computer calculation processing by summing up the calculation processing fees of each calculation processing unit until at least the next synchronization timing; ,
Is further executed by the computer processing for management,
6. The computer arithmetic processing division method according to claim 4, wherein the input step is executed when the second arithmetic processing fee is lower than the third arithmetic processing fee.   Information on the calculation processing time of the calculation processing unit executed by each computer is received from each computer, and the calculation processing time is multiplied by the sum of the unit price of energy consumed by the computer and the unit price of CPU time for each calculation processing unit. Calculating a calculation processing fee for each calculation processing unit, and calculating a calculation processing fee for the computer calculation processing by summing all the calculation processing fees for each calculation processing unit, The computer arithmetic processing dividing method according to claim 4, further executing by computer arithmetic processing.   A computer usage charge calculation program comprising a computer program for executing the charge calculation processing on a computer in the computer usage charge calculation apparatus according to claim 1 or 2.   The computer arithmetic processing division program which has a computer program which respectively performs the process of each said process in the computer arithmetic processing division | segmentation method of any one of Claims 4-7 on a computer.

Images