JP5151203B2 - Job scheduling apparatus and job scheduling method - Google Patents

Description

  The present invention relates to a job scheduling apparatus and a job scheduling method, and more particularly to a job scheduling apparatus and a job scheduling method for assigning a job to a computer.

  In recent years, with the progress of processing capacity per unit volume of computers, power consumption and heat generation per unit volume have been increasing. Recently, the processing performance of computers has been limited to the amount of heat generated by its own power consumption. This trend is the same in parallel computer systems composed of multiple computers, but in parallel computer systems where computers are often installed close to each other, there is a problem of heat beyond the heat countermeasures that are installed individually. The close correspondence of is demanded.

  As a countermeasure against heat in a parallel computer system, for example, even if each computer is operated at the maximum output, the computers should be connected in advance so that the computer will not stop or performance will deteriorate due to overheating. It is considered that the safest measure is to install it at some distance. However, considering the operation of computers in recent years, the situation where all computers are operated at the maximum output is often a rare case. If servers are arranged apart from each other in consideration of such operations, the computing capacity per unit volume of the entire system will be reduced. When performance is required in a limited volume, there is a need to place computers in close contact with each other and take measures against heat during operation. Also, in recent years, there is often a demand for operating a computer that generates a lot of heat in a machine room that does not have enough cooling capacity because the capital investment of the machine room is not in time for the increase in power consumption of the computer. In such cases, it may be necessary to solve the thermal problem through operation.

  As one form of such operation, there is a method of preventing overheating of a specific computer by distributing the computers that generate heat by devising the arrangement of jobs. However, there has not yet been proposed a job scheduling method for precisely arranging jobs while considering the influence of temperature on other computers due to job placement on a computer.

  As one of the flexible heat countermeasures in operation, there can be considered a method of controlling the temperature of a certain location so as not to be overheated by devising job scheduling to disperse the generation of heat. As one method, a job scheduling apparatus that monitors the temperature of each computer and always assigns a job to the computer having the lowest temperature has already been proposed (see, for example, Patent Document 1).

  Also, in a multiprocessor system, it is checked whether there is a processor that may cause the control IC to overheat, and if there is no possibility of overheating, the thread is dispatched according to the processor assignment, and the possibility of overheating There is also known a multiprocessor system that suppresses processing stoppage due to overheating by waiting for job assignment until a new free processor is created in addition to the current free processor (for example, , See Patent Document 2).

  Furthermore, a processor system that calculates a predicted temperature of a processor and assigns a task to the processor if the predicted temperature is within a range of conditions is also known (see, for example, Patent Document 3).

JP 2004-126968 A (page 6) JP 2006-099624 A (page 9) JP 2006-133955 A (pages 12-13)

  However, the conventional job scheduling apparatus described in Patent Document 1 does not consider the change in the ambient temperature due to job assignment, and the computer B near the computer A that was operating at the limit temperature has high power consumption. By assigning a job, it is conceivable that the ambient temperature of computer A rises and computer A goes down.

  In addition, in the inventions described in Patent Document 2 and Patent Document 3, temperature prediction is performed with reference to only a temporary amount of heat generated by assigning a task, so accurate temperature prediction cannot be performed and the computer is down. There is a high risk of Furthermore, in any of the inventions described in Patent Documents 1 to 3, the temperature is predicted using a secondary temperature change caused by the amount of heat generated by job assignment to computers and the surrounding computers, or the amount of heat generated in the surrounding computers. Because the temperature is not taken into consideration, it is not possible to bring out higher performance within the range where the computer does not go down.

  The present invention has been made in view of the above points, and a job scheduling apparatus and job scheduling capable of preventing a computer performance down and a system down due to overheating of the system by performing temperature prediction in consideration of the calorific value of surrounding computers. It aims to provide a method.

  Another object of the present invention is to provide a job scheduling apparatus and a job scheduling method capable of extracting the maximum capacity within a predetermined temperature limit even when a computer is operated in a machine room having insufficient cooling capacity. Is to provide.

To achieve the above object, the job scheduling apparatus of the present invention is connected via a network to a plurality of computers, a job scheduling apparatus for allocating jobs for each computer in advance a plurality of all jobs assignable A job history table that stores, for each type, a history that associates the power consumption of the computer to which each job is assigned and the execution time of the job. Job classification means that outputs the classification result of the type according to the job, and average power consumption and average job execution for the desired job type obtained by referring to the job history table of each computer based on the classification result of the job classification means When a desired job is assigned to each computer of multiple computers based on time, it is deleted for each computer. Predicting the amount of power increase, from the predicted value of the power consumption increase, and the internal temperature information and power consumption information for at least each computer which is detected for each computer to be input from each of the plurality of computers, when assigned a desired job on one of the computer of the plurality of computers, performs the one computer and one computer of the plurality of predicting the respective temperature changes in the calculation device near the computer each computer of An assignment determination unit that determines a computer to which a desired job is assigned based on the obtained predicted temperature change value of each computer; a job assignment unit that assigns a desired job to the computer determined by the assignment determination unit; Get the power consumption information after a certain period of time for the computer to which the job is assigned, and use the job type as an index after the job is completed. And having a job history table managing means for adding updates the value of the power consumption during job execution in the job history table with.

To achieve the above object, the job scheduling apparatus of the present invention, save each size information, environment information for each computer based on the computer installation location information and location cooling capability information of the multiple computers And a plurality of computers further have an environmental temperature measuring means for measuring the environmental temperature,
The allocation determining means is a sum of a predicted value of the temperature change of each of the job allocation computer which is one computer to which a desired job is allocated and a computer around the job allocation computer, and the environmental temperature from the environmental temperature measurement means. The predicted temperature of each computer does not exceed the maximum allowable environmental temperature set based on the environmental information read from the environmental information storage device in each of the job allocation computer and the computers surrounding the job allocation computer, and the predicted temperature The job assignment calculator when the difference between the maximum allowable environmental temperature and the maximum allowable environment temperature indicates the minimum value and the maximum difference temperature among the difference temperatures indicating the minimum value is determined as the computer to which the job is allocated. When there is no computer indicating the difference temperature, job assignment is suspended .

  In the present invention, when the cooling capacity of the place (machine room) where the computer is installed is insufficient, a computer suitable for job allocation can be obtained within a predetermined temperature limit (maximum allowable environment temperature). . In this case, the highest performance of the computer system cannot be extracted, but the investment in the machine room is intentionally reduced by looking at the balance between cost and performance, or the latest with high computing power before the machine room facilities are ready. It is possible to arrange the equipment later by operating the computer.

  Further, according to the present invention, when a computer satisfying the condition is not found as a result of prediction, job assignment is suspended and the currently assigned job is awaited. As a result, it is possible to avoid system down and performance down due to temperature problems.

In order to achieve the above object, according to the job scheduling apparatus of the present invention, the allocation determining means includes at least temperature information inside each computer detected for each computer input from each of the plurality of computers, and Based on the power consumption information, the temperature difference between each computer before and after assigning the desired job to each computer was measured, and the increase in power consumption indicated by the power consumption information was measured for each computer . The correlation function is obtained from the correlation with the temperature rise value indicated by the temperature difference, and the temperature rise value in the surrounding computers is estimated based on the power consumption increase value of each computer using the correlation function , and the estimation It is a means for determining a computer to which a desired job is assigned based on a value.

In order to achieve the above object, the job scheduling method of the present invention is a job scheduling method for allocating jobs to each computer connected to a plurality of computers via a network, and assigning all assignable jobs. There is a job history table that stores the history that associates the power consumption of the computer and the execution time of the job that has been classified into a plurality of types in advance and assigned a job to each of the plurality of computers, and tries to assign it. A first step for outputting a classification result of a type corresponding to a desired job, and an average power consumption of the desired job type obtained by referring to the job history table of each computer based on the classification result of the first step Each computer when a desired job is assigned to each computer of multiple computers from the average job execution time Of a second step of predicting a power consumption increase, and the predicted value of power consumption increase by a second step, at least the computer is detected for each computer to be input from each of the plurality of computers from the internal temperature information and power information, predicts the respective temperature changes in the desired job when assigned to one computer of the plurality of computers, one computer and calculation unit near one of the computer A third step of determining a computer to which a desired job is assigned based on the obtained predicted value of temperature change of each computer, and a computer determined by the third step respect, a fourth step of allocating a desired job, obtains power consumption information after a certain time of a computer assigned a job, the job after the job is completed S as an index to, characterized in that it comprises a fifth step of adding update the power consumption value of the time and during job execution to the job completion to the job history table.

  In the present invention, the job scheduling apparatus has information such as the installation position of each computer and the surrounding environment, predicts the temperature change around each computer due to job input and movement by numerical calculation, and assigns jobs based on the prediction. The computer to perform can be determined. In order to predict the temperature change around each computer, it is necessary to predict the amount of heat generated when a job is assigned in advance. In the present invention, a method is used in which a history of power consumption increase when a job is actually allocated is retained and analogized there. Jobs are classified into several types in advance, and a history is taken for each type. The predicted value of power consumption can be obtained with a certain degree of accuracy from the power consumption value before allocating the job and the average value of the history information.

  According to the present invention, by predicting temperature changes of the own computer and all peripheral computers when a job is assigned to each computer, and determining a computer to which a desired job is assigned based on the predicted value, In addition to the calorific value generated as a result of assigning, it is possible to determine the job assignment by predicting the temperature including the calorific value of surrounding computers, and overheating with a specific computer among the multiple computers that make up the system Can be prevented, system down and overall performance degradation can be prevented.

  In addition, according to the present invention, even when a computer is operated in a machine room with insufficient cooling capacity, the maximum capacity can be extracted within a predetermined temperature limit. If it is not found, it is possible to avoid a system down or a performance down due to a temperature problem by suspending the job assignment and waiting for the currently assigned job to end.

  Next, the best mode for carrying out the present invention will be described with reference to the drawings. FIG. 1 shows a configuration diagram of an embodiment of a job scheduling apparatus according to the present invention. In the figure, a parallel computer system 100 is a computer system in which computers 111, 112, 113, etc. are connected by a network 101. Job assignment in the parallel computer system 100 is performed by the scheduling device 103 connected to the same network 101. Each of the computers 111, 112, and 113 includes an environmental temperature sensor 121, a CPU temperature sensor 122, and a power consumption sensor 123, and the scheduling device 103 can read out such information through the network 101. Each computer 111, 112, 113 is set with a maximum allowable environmental temperature 124, and can be read from the scheduling device 103 in the same manner. This maximum allowable environmental temperature information is set in advance in a nonvolatile memory, for example. As another implementation, the maximum allowable environmental temperature 124 can be rewritten from the outside and set according to the cooling capacity of the machine room.

  In the present embodiment, each computer 111, 112, 113 automatically adjusts the cooling temperature in the computer according to the value of the environmental temperature sensor 121 and the value of the CPU temperature sensor 122. This cooling control is performed by, for example, a BMC (Baseboard Management Controller). Due to this function, the present embodiment does not consider cooling of the computers 111, 112, and 113.

  The network 101 is connected to an environment information storage device 102 for storing location information where the computers 111, 112, and 113 are installed. This environment information can also be read from the scheduling device 103. As another mounting method, mounting in which the position information is stored in each of the computers 111, 112, and 113 is also conceivable. In addition, the environmental temperature storage device 102 includes size information (width, depth, height) of each computer 111, 112, 113 and data 111, 112, 113 as data for more accurate temperature prediction. Stores the structural information of the installed machine room and the cooling capacity of the machine room. The scheduling apparatus 103 can accurately perform temperature prediction by reading out this information and using it as a boundary condition in the calculation.

  FIG. 2 shows a configuration diagram of an example of the scheduling apparatus 103. In the figure, the assignment determination unit 203 assigns jobs among computers based on the temperature sensor values, power consumption sensor values, estimated power consumption values when jobs are assigned, environmental parameters, and the like. The temperature change is predicted, and the job assignment destination computer is determined based on the predicted value. A job classifier 201, a job history table 202, and a history table management unit 204 exist for predicting the power consumption of a job.

  Since it is difficult to predict the increased power consumption for all jobs, the jobs are classified in advance into a plurality of types, and the job history table 202 stores a history for each type. The job types in this case are, for example, database search transactions, web (Web) requests, data processing, and the like, and the scheduling apparatus 103 performs classification.

  FIG. 3 shows an example of the job history table 202. As shown in the figure, the job history table 202 stores history, power consumption increase, and job time in association with each job type of Web request, database search, data processing 1, and data processing 2. . The job history table 202 is provided corresponding to each of a plurality of computers 111, 112, 113 for which the scheduling device 103 connected to the network 101 in FIG. Therefore, the job history table 202 shown in FIG. 3 is a job history table for one computer connected to the network, and the “history” therein is the history of the one computer.

  Next, the operation of the present embodiment will be described in detail. First, installation position information of each of the computers 111, 112, and 113 is input in advance to the environment information storage unit 102 in FIG. The installation position information may be simple information such as a vertical direction, a horizontal direction, and a height direction in the machine room. Although it is assumed that the installation position information is input manually, means for automatically inputting the installation position information of each of the computers 111, 112, and 113 using a known technique such as RFID is provided. Also good. Similarly, various parameters necessary for calculating the environmental temperature are set in the environmental information storage device 102 in FIG. This is, for example, information such as size information of each computer 111, 112, 113, machine room structure information, heat exhaust direction of each computer 111, 112, 113, and heat exhaust capability of the entire machine room.

  FIGS. 4A and 4B are a configuration diagram and a flowchart for explaining a method of determining a computer to which a job is assigned. As shown in FIGS. 4A and 4B, the job classifier 201 classifies the job (step S1), and the allocation determining unit 203 increases the power consumption when the job is allocated from the job history table 202. Get the expected value. At the same time, the value of the environmental temperature sensor 121 of each computer 111, 112, 113 and the environmental parameter of each computer 111, 112, 113 are acquired (step S2).

  Subsequently, the assignment determination unit 203 determines the assignment destination by predicting the environmental temperature after a predetermined time in all the computers 111, 112, 113 when the job is assigned to each computer (step S3). Thereafter, the job classifier 201 assigns a job to the assignment destination (step S4).

  The above job assignment will be specifically described. The scheduling apparatus 103 estimates the increase in power consumption W_N when a certain job is assigned to a certain computer N, for example, based on the average of the history for each job type in FIG. If FIG. 3 is the job history table of the computer 111 of FIG. 1, for example, the increase in the average power consumption when each job is assigned to the computer 111 is as shown in Table 1 below from the average power consumption and the average time. Is required.

これを各計算機について計算し、その消費電力量に基づいて、ジョブの割当先の計算機を決定する。

This is calculated for each computer, and the computer to which the job is assigned is determined based on the power consumption.

Next, the method for determining the job assignment destination in step S4 will be described more specifically. First, based on the increase in power consumption Ws (N, K) when job K is assigned to a certain computer N , the own computer (here, computer N) that is the one computer to which job K is assigned and the surroundings Estimate the temperature rise of the computer. When the job K is assigned to the computer N, the estimated temperature rise of the computer N or the surrounding computer M is Tup (M, N, K), and the upper limit environmental temperature Tlim (M) is set for all the surrounding computers M. A search is made for combinations of N and K that do not exceed, and allocation is performed.

  For example, the computers 111, 112, and 113 in FIG. 1 are expressed as computers 1, 2, and 3, and the environmental temperature before assigning jobs of these computers is shown in Table 2, and Assume that the upper limit environmental temperature is 40 ° C. and the average power consumption of each computer is as shown in Table 3 when a Web request job is assigned to each computer.

次に、その平均消費電力量を用いて計算機ｘにジョブを割り当てた時の、計算機（自計算機と他計算機）ｙの温度上昇量の推定値を表４のように求め、計算機ｘにジョブを割り当てた時の、計算機（自計算機と他計算機）ｙの温度の推定値を表５のように求める。

Next, when the job is assigned to the computer x using the average power consumption, an estimated value of the temperature rise amount of the computer (own computer and other computers) y is obtained as shown in Table 4, and the job is assigned to the computer x. The estimated value of the temperature of the computer (own computer and other computer) y at the time of allocation is obtained as shown in Table 5.

ただし、表４及び表５において縦軸の計算機は上記の計算機ｘを示し、横軸の計算機は上記の計算機ｙを示す。従って、例えば、表４の２行１列目の「０.３」は計算機２（１１２）にジョブを割り当てた時の、計算機１（１１１）の温度上昇の推定値を示し、表５の２行１列目の「37.5」は計算機２（１１２）にジョブを割り当てた時の、計算機１（１１１）の温度の推定値を示す。

However, in Tables 4 and 5, the vertical axis computer indicates the above-mentioned computer x, and the horizontal axis computer indicates the above-mentioned computer y. Therefore, for example, “0.3” in the second row and first column of Table 4 indicates an estimated value of the temperature rise of the computer 1 (111) when a job is assigned to the computer 2 (112). “37.5” in the first column of the row indicates an estimated value of the temperature of the computer 1 (111) when a job is assigned to the computer 2 (112).

  As a result, when a job is assigned to computer 1 (111) from Table 5, the computer y having the smallest difference between the upper limit environmental temperature (40 ° C.) and the estimated temperature is 1.2 ° C. of computer 2 (112). Yes, and when the job is assigned to the computer 2 (112), the computer y having the smallest difference between the upper limit environment temperature (40 ° C.) and the estimated temperature value is 0.2 ° C. of the computer 2 (112). When a job is assigned to the computer 3 (113), the computer y having the smallest difference between the upper limit environmental temperature (40 ° C.) and the estimated temperature is 1.25 ° C. of the computer 2 (112). . Of these, 1.25 ° C. having the largest differential temperature is a case where the job is assigned to the computer 3 (113), and therefore, the margin from the upper limit environmental temperature (40 ° C.) is the largest to be assigned to the computer 3 (113). Therefore, the scheduling apparatus 103 assigns a job to the computer 3 (113).

  Next, management of the job management table 202 will be described with reference to the configuration diagram of the scheduling apparatus 103 in FIG. 5A and the flowchart in FIG. First, the job classifier 201 classifies jobs into certain types in advance (step S11). Subsequently, the job classifier 201 determines a job assignment destination computer (step S12), and after the job is input to each computer (step S13), the history table management unit 204 uses the power consumption sensor of the input computer after a predetermined time. The value of 123 is read (step S14).

  Subsequently, when the history table management unit 204 confirms the completion of the job of the computer, the job type uses the job type as an index to calculate the time until completion of the job and the power consumption value detected by the power consumption sensor 123 during job execution. Additional writing is performed in the history table 202 (step S15). Since the job history table 202 can have only a finite entry for each job type, the entry is overwritten by the round robin method.

  By the way, as a method for obtaining an estimated value of the temperature rise of another computer from the power consumption rise value, there are the following methods. The scheduling device 103 first takes a power consumption history as shown in FIG. 3 every time a job is assigned, and obtains the power consumption amount from this, and then calculates the temperature difference between all computers before the job is assigned and when the job is finished. Take. Subsequently, the scheduling apparatus 103 creates a correlation diagram as shown in FIG. 6 of the increase value of power consumption and the increase value of temperature for each computer, and obtains a correlation function based on the correlation diagram (for example, by a linear expression). . Using this correlation function, the scheduling device 103 estimates, for each computer, the temperature increase value in another computer from the increase value of the power consumption of that computer.

  The environmental temperature can be predicted by solving the thermal diffusion by numerical calculation. In this case, the scheduling device 103 reads information necessary for this from the environment information storage device 102, sets an appropriately sized grid, and numerically calculates the time evolution of temperature change due to heat conduction or convection. When it is determined that the system has reached a certain level of equilibrium, the numerical calculation is stopped, and the predicted value of the temperature change in each of the other computers at that time is stored.

  The scheduling apparatus 103 makes this prediction for all the computers 111, 112, and 113, and employs the job assignment in which the temperature condition becomes the most lenient. The job with the mildest temperature condition is an assignment in which the difference between the maximum allowable environment temperature and the environment temperature of the computer with the severest temperature condition (the computer with the smallest difference between the maximum allowable environment temperature and the environment temperature) is the largest. . In other words, assuming that the predicted environmental temperature in computer j (including its own computer i) when assigning a job to computer i is Tij, and the maximum allowable environmental temperature of computer j is TCj, the job is assigned to each computer i. Among the plurality of computers j having the smallest difference between the maximum allowable environment temperature TCj and the predicted environment temperature Tij in each computer j at the time, the computer i to which the job indicating the maximum difference temperature is assigned, that is, , Max_i (min_j (TCj-Tij)) is assigned. This computer i becomes a job assignment destination computer whose temperature condition is the least severe.

  When such a computer i does not exist, the scheduling apparatus 103 waits for the job assignment, waits for any currently executed job to end, and determines the assignment destination again. In addition, the maximum allowable environmental temperature of each computer can be rewritten, and when the cooling capacity of the machine room is insufficient, this value is adjusted to prevent overheating exceeding the cooling capacity of the machine room. You can also. The set value TC_limit in this case is obtained in advance when the computer is installed.

It is a block diagram of one embodiment of a job scheduling apparatus of the present invention. It is a block diagram of an example of the scheduling apparatus in FIG. FIG. 3 is a diagram illustrating an example of a job history table in FIG. 2. It is a block diagram and a flowchart for explaining operation of a scheduling device for explaining a method of determining a computer to which a job is assigned. It is a block diagram of a scheduling apparatus for demonstrating the management method of a job management table, and the flowchart for operation | movement description. It is a figure which shows an example of correlation with the increase in power consumption of a computer, and a temperature rise.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Parallel computer system 101 Network 102 Environment information storage apparatus 103 Scheduling apparatus 111, 112, 113 Computer 121 Environment temperature sensor 122 CPU temperature sensor 123 Power consumption sensor 124 Maximum permissible environment temperature 201 Job classifier 202 Job history table 203 Assignment determination part 204 History table management department

Claims (6)

A job scheduling apparatus that is connected to a plurality of computers via a network and assigns a job to each computer,
Job history in which all assignable jobs are classified into a plurality of types in advance, and a history in which the power consumption of the computer to which the job is assigned and the execution time of the computer is associated with each of the plurality of computers is stored for each type A job classification unit having a table and outputting a classification result of a type corresponding to a desired job to be allocated ;
From the average power consumption and average job execution time of the desired job type obtained by referring to the job history table of each computer based on the classification result of the job classification means, the desired job is assigned to the plurality of computers. Predicts the increase in power consumption for each computer when it is assigned to each computer separately, and detects the predicted value for the increase in power consumption and each computer input from each of the plurality of computers From at least the temperature information and power consumption information inside each of the computers , when the desired job is assigned to one computer of the plurality of computers, the one computer and the periphery of the one computer conducted for calculation machine for each of said plurality to predict the temperature change computing the calculator, based on the predicted value of the temperature change of each computer obtained the desired And allocation determining means for determining a computer to assign a job,
Job assignment means for assigning the desired job to the computer determined by the assignment determination means ;
Obtains power consumption information after a certain time of a computer to which a job is assigned, and uses the job type as an index after completion of the job, and the job history table shows the time until job completion and the value of power consumption during job execution. And a job history table management means for additionally updating the job scheduling apparatus. Wherein each of the size information of the multiple computers, and has a computer installation position information and the environment information storing device based on the cooling capacity information of the installation location to store the environmental information for each computer, the plurality of computers under ambient It further has an environmental temperature measuring means for measuring,
The allocation determining means includes
The predicted value of the temperature change of each of the job allocation computer that is the one computer to which the desired job is allocated and computers around the job allocation computer, and the sum of the environmental temperature from the environmental temperature measurement means The predicted temperature of each computer does not exceed the maximum allowable environment temperature set based on the environment information read from the environment information storage device in each of the job allocation computer and the computers around the job allocation computer, A job assignment computer is assigned when the temperature of the difference between the predicted temperature and the maximum allowable environmental temperature indicates a minimum value, and the temperature of the difference of the difference indicating the minimum value indicates the maximum temperature. determined as a computer, when the computer indicative of the temperature of maximum difference is not present, to characterized in that pending job assignments Job scheduling apparatus according to claim 1.   The allocation determining means is configured to allocate the desired job to each computer based on at least temperature information and power consumption information inside each computer detected for each computer input from each of the plurality of computers. The correlation function is calculated from the correlation between the increase value of the power consumption indicated by the power consumption information and the increase value of the temperature indicated by the measured temperature difference for each computer. A means for estimating a temperature increase value in a peripheral computer based on an increase value of power consumption of each computer using the correlation function, and determining a computer to which the desired job is assigned based on the estimated value The job scheduling apparatus according to claim 1, wherein the job scheduling apparatus is provided. A job scheduling method that is connected to a plurality of computers via a network and assigns a job to each computer,
Job history in which all assignable jobs are classified into a plurality of types in advance, and a history in which the power consumption of the computer to which the job is assigned and the execution time of the computer is associated with each of the plurality of computers is stored for each type A first step having a table and outputting a classification result of a type corresponding to a desired job to be assigned ;
From the average power consumption and average job execution time in the desired job type obtained by referring to the job history table of each computer based on the classification result of the first step, the desired job is assigned to the plurality of jobs. A second step of predicting an increase in power consumption for each computer when each computer is assigned separately;
From the predicted value of the increase in power consumption by the second step, and at least the temperature information and power consumption information inside each computer detected for each computer input from each of the plurality of computers , when assigned a desired job on one of the computer of the plurality of computers, the one computer and the one computer of said plurality to predict the respective temperature changes in the calculation device near the computer A third step that is performed for each computer and determines a computer to which the desired job is assigned based on the obtained predicted value of the temperature change of each computer ;
A fourth step of assigning the desired job to the computer determined in the third step ;
Obtains power consumption information after a certain time of a computer to which a job is assigned, and uses the job type as an index after completion of the job, and the job history table shows the time until job completion and the value of power consumption during job execution. And a fifth step of additionally updating the job scheduling method. Wherein each of the size information of the multiple computers, and has a computer installation position information and the environment information storing device based on the cooling capacity information of the installation location to store the environmental information for each computer, the plurality of computers under ambient It further has an environmental temperature measuring means for measuring,
The third step includes
The predicted value of the temperature change of each of the job allocation computer that is the one computer to which the desired job is allocated and computers around the job allocation computer, and the sum of the environmental temperature from the environmental temperature measurement means The predicted temperature of each computer does not exceed the maximum allowable environment temperature set based on the environment information read from the environment information storage device in each of the job allocation computer and the computers around the job allocation computer, A job assignment computer is assigned when the temperature of the difference between the predicted temperature and the maximum allowable environmental temperature indicates a minimum value, and the temperature of the difference of the difference indicating the minimum value indicates the maximum temperature. determined as a computer, when the computer indicative of the temperature of maximum difference is not present, to characterized in that pending job assignments Job scheduling method of claim 4. The third step assigns the desired job to each computer based on at least temperature information and power consumption information inside each computer detected for each computer input from each of the plurality of computers. Measure the difference in temperature of each computer before and after allocation, and correlate from the correlation between the increase value of power consumption indicated by the power consumption information for each computer and the increase value of temperature indicated by the difference in measured temperature A temperature rise value estimating step for obtaining a function, and estimating a temperature rise value in a surrounding computer based on an increase value of power consumption of each computer using the correlation function, and based on the estimated value of the temperature rise value 5. The job scheduling method according to claim 4 , further comprising a determination step of determining a computer to which a desired job is assigned.

Images