JP7331386B2 - Job control system, method and program - Google Patents

Description

TECHNICAL FIELD The present disclosure relates to a job control system, method, and program, and more particularly to a job control system, method, and program that control execution of jobs on input data.

As a general batch processing operation, a method of sequentially operating batch processing based on a schedule using a job management system or the like is known. In the operation of such batch processing, if processing is performed using a fixed resource, the processing time varies due to differences in the amount of input data on a daily basis. In the operation of batch processing, even if the amount of input data fluctuates, it is important to secure the necessary resources and complete the processing by the scheduled end time.

As a related technique, Patent Literature 1 discloses a batch control system. In the batch control system described in Patent Document 1, the monitoring unit monitors the resource usage and processing time when the first virtual computer executes the first batch processing, and obtains the maximum resource usage and maximum processing time. . The prediction unit predicts the resource usage and processing time when the first virtual computer executes the second batch processing from the tendency of the maximum resource usage and maximum processing time. The control unit controls execution of the second batch process according to the predicted resource usage and the predicted processing time.

As another related technique, Patent Literature 2 discloses a job execution method. In Patent Document 2, a job that can be divided into an arbitrary number of tasks is executed using a plurality of computers connected to a network. In Patent Document 2, one computer among a plurality of computers processes a task of a job for a predetermined period of time. The computer calculates the progress rate of task processing in a predetermined time, and estimates the completion time of task processing in one computer based on the progress rate and the predetermined time.

The computer determines whether the estimated completion time satisfies the job requirements. For example, when the remaining time until the deadline time (desired execution end time) of the job is longer than the estimated completion time, the computer determines that the requirements are not satisfied. If the requirements are met, the remaining tasks are performed using one computer. If the requirements are met, move some of the tasks to other computers and have the other computers perform some of the tasks.

JP 2017-129988 A Japanese Patent Application Laid-Open No. 2008-123205

In recent years, the use of cloud computing has spread, and batch processing has also been increasingly performed on the cloud. When performing batch processing on cloud computing, the more resources are allocated, the higher the cost. Therefore, in order to keep costs down, it is important to allocate the necessary amount of resources without waste.

However, Patent Document 1 assumes resource management on a physical server. In Patent Literature 1, the maximum value or the average value always derived from the past record is used as the resource usage rate as the predicted value. Batch processing does not always show a simple increase or decrease in processing time, and the amount of input data for batch processing may vary from day to day. Therefore, the required amount of resources may vary depending on the day. In the system described in Patent Document 1, for example, if scale adjustment is performed using the maximum value as a predicted value, there is a problem that wasteful costs are incurred when the amount of input data is small. In addition, when the average value is used as the predicted value, when the amount of input data is large, the amount of resources used exceeds the amount of allocated resources, and there is a problem that processing cannot be completed by the scheduled end time.

With respect to US Pat. No. 6,200,000, if the estimated completion time satisfies the requirements of the job, all remaining tasks are executed on one computer. In this case, the amount of resources may be excessive for the number of tasks. For this reason, when the amount of data is small in batch processing, it is not possible to allocate the required amount of resources without waste, and the cost cannot be reduced.

In view of the circumstances described above, the present disclosure aims to provide a job control system, method, and program capable of finishing processing within time while keeping costs down.

In order to achieve the above object, the present disclosure provides job execution control means for causing a virtual machine configured such that two or more resource allocation amounts can be changed to execute a job for input data, and a combination of the resource allocation amounts. Based on the execution result data including the processing time for the input data in a predetermined unit, the data amount of the input data, and the scheduled end time, the job is completed by the scheduled end time and the resource allocation amount is reached. allocation resource prediction means for predicting a combination of the resource allocation amounts that minimizes the cost determined based on the combination, wherein the job execution control means adjusts the resource allocation amount of the virtual machine according to the predicted resource allocation amount; A job control system is provided that controls the combination of quotas.

The present disclosure provides execution result data including a combination of the resource allocation amounts in a virtual machine configured such that the allocation amount of two or more resources can be changed, and a processing time for input data in a predetermined unit, and execution in the virtual machine. Based on the amount of input data to be processed by the job and the scheduled end time, the job will be completed by the scheduled end time, and the cost determined based on the combination of the resource allocation amounts will be minimized. A job control method for predicting a combination of the resource allocation amounts, controlling the resource allocation amount of the virtual machine to the predicted combination of the resource allocation amounts, and causing the virtual machine to execute a job for the input data. I will provide a.

The present disclosure provides execution result data including a combination of the resource allocation amounts in a virtual machine configured such that the allocation amount of two or more resources can be changed, and a processing time for input data in a predetermined unit, and execution in the virtual machine. Based on the amount of input data to be processed by the job and the scheduled end time, the job will be completed by the scheduled end time, and the cost determined based on the combination of the resource allocation amounts will be minimized. A process for predicting a combination of the resource allocation amounts, controlling the resource allocation amount of the virtual machine to the predicted combination of the resource allocation amounts, and causing the virtual machine to execute a job on the input data. provides a program for causing a computer to execute

A job control system, method, and program according to the present disclosure can control a job so as to finish processing within a time while keeping costs down.

1 is a schematic block diagram of a job control system according to the present disclosure; FIG. 1 is a block diagram showing an entire system including a job control system according to an embodiment of the present disclosure; FIG. FIG. 10 is a diagram showing an example of settable values and costs of target resources; The figure which shows an example of input data. 4 is a flowchart showing the flow of overall processing; 4 is a flowchart showing a procedure for predicting the amount of allocated resources; 4 is a flowchart showing a procedure for predicting the amount of allocated resources; FIG. 11 is a diagram showing an example of a process of calculating a combination of resource allocation amounts in binary exploration; FIG. 10 is a diagram showing a process of calculating a combination of allocation amounts of minimum-cost resources; (a) to (d) are schematic diagrams showing the process of searching for a combination of resource allocation amounts with the lowest cost. FIG. 2 is a block diagram showing a configuration example of an information processing device;

An outline of the present disclosure will be described prior to description of the embodiments of the present disclosure. FIG. 1 schematically illustrates a job control system according to the present disclosure. The job control system 10 has job execution control means 11 and resource allocation prediction means 12 . The job execution control means 11 causes the virtual machine 20 to execute a job for input data. The virtual machine 20 is configured such that two or more resource allocation amounts can be changed. In the virtual machine 20, the cost is determined based on a combination of resource allocation amounts.

The execution result data 15 includes a combination of resource allocation amounts in the virtual machine 20 and a processing time for a predetermined unit amount of input data. Based on the execution result data 15, the amount of input data, and the scheduled end time, the resource allocation prediction unit 12 predicts a combination of resource allocation amounts that will result in the job ending by the scheduled end time and with the lowest cost. do. The job execution control unit 11 controls the resource allocation amount of the virtual machine to a combination of predicted resource allocation amounts.

In this embodiment, the execution result data 15 includes a combination of resource allocation amounts and a processing time for a predetermined unit amount of input data. By referring to the execution result data 15, the resource allocation prediction means 12 can estimate how much time is required to process the input data for each combination of resource allocation amounts. The allocated resource prediction means 12 predicts the combination with the lowest cost among the combinations of resource allocation amounts that can finish the processing of the input data by the scheduled end time. The job execution control unit 11 controls the resource allocation amount of the virtual machine 20 to the predicted combination, and executes the job for the input data. In this way, when cloud computing or the like is used, efficient resource allocation to the virtual machines 20 can be realized, and job execution costs can be suppressed.

Embodiments of the present disclosure will be described below with reference to the drawings. FIG. 2 shows an overall system including a job control system according to one embodiment of the present disclosure. System 100 has a job control system 110 and a job execution system 150 . The job control system 110 controls execution of jobs such as batch processing. The job execution system 150 executes job processes. In system 100, job control system 110 may be configured as a manager and job execution system 150 may be configured as an agent.

The job execution system 150 includes input data 151 , a job process 152 and an OS (Operating System) 153 . The job execution system 150 is configured as a virtual machine or container on the virtualization infrastructure 200 . The resources allocated to the job execution system 150 in the virtualization platform 200 are configured to be changeable by the virtualization platform 160 while the job execution system 150 is in operation. The cost of the job execution system 150 is determined according to the allocated amount of each resource. A job execution system 150 corresponds to the virtual machine 20 in FIG.

FIG. 3 shows an example of configurable values and costs for target resources. Here, CPUs (Central Processing Units) and disks are considered as target resources. It is assumed that the CPU can allocate resources in five stages: lowest speed (1), low speed (2), normal (3), high speed (4), and highest speed (5). Assume that the maximum cost is 100 when the CPU resource allocation (setting value) is the highest speed (5), and the minimum cost is 10 when the lowest speed (1). It is also assumed that the disk can be allocated resources in five stages: lowest speed (1), low speed (2), normal (3), high speed (4), and highest speed (5). The fastest disk resource allocation (5) has a maximum cost of 30, and the slowest (1) has a minimum cost of 10.

For example, if the resource allocation for the CPU is the fastest (5) and the resource allocation for the disk is the fastest (5), the maximum cost is 130. On the other hand, if the resource allocation for the CPU is the slowest (1) and the resource allocation for the disk is the slowest (1), the cost is the lowest 20. The time required to process a job varies depending on the combination of CPU and disk setting values. If the allocation of resources to each resource is the maximum, the cost is the highest, but the processing can be completed in a short time. On the other hand, if the allocation of resources to each resource is minimal, the cost is the lowest, but the processing time is long.

In job execution system 150 , input data 151 is given to job process 152 . The input data 151 is data that can be processed in units of records or the like. The job process 152 executes predetermined processing on the input data 151 on a record-by-record basis. A record unit corresponds to the minimum unit in which the job process 152 executes processing. In the job execution system 150, it is assumed that the job process 152 activated at one time is a single job process.

The job control system 110 has a job definition data storage unit 111 , an input data confirmation unit 112 , a job execution control unit 113 , an allocation resource prediction unit 114 and an execution result data storage unit 115 . A job control system 110 corresponds to the job control system 10 in FIG.

The job definition data storage unit 111 stores definition information (job definition data) on jobs. At least part of the content of the job definition data is predefined by the user. Job definition data includes, for example, batch processing to be executed, input data information, and scheduled end time. The job definition data further includes the ID of the job definition data for retrieving past results generated by the system, and information as to whether or not the batch process to be executed has been changed since the last execution. .

The input data information includes, for example, a command for confirming the total amount of the input data 151, a command for obtaining the next minimum unit of input data based on the last processed data, and a command for obtaining the remaining input data based on the last processed data. Contains get commands. As for the input data information, it is expected that the commands will differ according to the input data, so the user shall specify each command in the job definition data.

FIG. 4 shows an example of input data. In this example, the input data has fields of "id", "name", "department", "gender", and "age". As an example, when a MySQL table as shown in FIG. 4 is used as the input data 151, the following SQL commands are defined as input data information in the job definition data.
An example of a check command for the total amount of input data:
mysql -u user -p passwd db -e 'SELECT COUNT(*) TABLE1'
An example of the next granular input data acquisition command based on the last processed data:
mysql -u user -p passwd db -e 'SELECT * FROM TABLE1 LIMIT 1 OFFSET <Number of last processed data>'
Example command to get remaining input data based on last processed data:
mysql -u user -p passwd db -e 'SELECT * FROM TABLE1 OFFSET <Number of last processed data>'

Returning to FIG. 2, the input data confirmation unit 112 confirms the total amount of input data 151 of the job. The job execution control unit 113 controls jobs executed in the job process 152 . In addition, the job execution control unit 113 requests the virtualization base 160 to change resource allocation of the job execution system 150 . A job execution control unit 113 corresponds to the job execution control unit 11 in FIG.

The execution result data storage unit 115 stores job execution results (execution result data) including information such as the allocation amount of each resource and the processing time per unit input. The resource allocation prediction unit 114 predicts the amount of resources allocated to the job execution system 150 . More specifically, based on the execution result data, the data amount of the input data 151, and the scheduled end time, the allocation resource prediction unit 114 allocates resources that will complete processing by the scheduled end time and that will minimize the cost. Predict quantity combinations. The execution result data stored in the execution result data storage unit 115 corresponds to the execution result data 15 in FIG. The allocated resource prediction unit 114 corresponds to the allocated resource prediction unit 12 in FIG.

The execution result data storage unit 115 stores execution result data when the job execution system 150 is caused to execute a job for minimum unit input data with respect to a combination of a plurality of resource allocation amounts. The allocation resource prediction unit 114 uses the execution result data stored in the execution result data storage unit 115 to determine whether or not the job will be finished by the scheduled end time for some combinations of resource allocation amounts. do.

The allocation resource prediction unit 114 reads out the execution result data corresponding to the combination of the corresponding resource allocation amounts from the execution result data storage unit 115 . The allocation resource prediction unit 114 calculates the remaining required processing time based on the processing time included in the read execution result data and the amount of remaining input data. The allocation resource prediction unit 114 compares the required time with the remaining time determined based on the scheduled end time, and determines whether or not the process will end by the scheduled end time. The allocated resource prediction unit 114 predicts a combination with the lowest cost among combinations of resource allocation amounts for which processing will be completed by the scheduled end time. The job execution control unit 113 controls the resource allocation amount of the job execution system 150 to a combination of resource allocation amounts predicted by the allocation resource prediction unit 114 .

FIG. 5 shows an overall flow of processing (job control method) in the job control system. The job execution control unit 113 reads job definition data from the job definition data storage unit 111 (step A1). The job execution control unit 113 requests the input data confirmation unit 112 to confirm the input data (step A2). In step A<b>2 , the job execution control unit 113 passes a confirmation command for the total amount of input data included in the input data information of the job definition data to the input data confirmation unit 112 . The input data confirmation unit 112 confirms the total amount of the input data 151 using the received confirmation command, and returns the total amount of the input data 151 to the job execution control unit 113 .

The job execution control unit 113 requests the resource allocation prediction unit 114 to predict resource allocation. The allocated resource prediction unit 114 predicts the allocated resource (step A3). The resource allocation prediction unit 114 predicts a combination of resource allocation amounts that allows processing to be completed by the scheduled time at the minimum cost, and returns the resource allocation prediction result (combination of the allocation amount of each resource) to the job execution control unit 113. do. The allocated resource prediction unit 114 may request the job execution control unit 113 to acquire execution result data in the allocation resource prediction. At that time, the allocation resource prediction unit 114 designates a combination of the allocation amount of each resource of the job execution system 150 and requests acquisition of execution result data. A detailed operation of the allocation resource prediction unit 114 in step A3 will be described later.

The job execution control unit 113 determines whether or not the allocation resource prediction unit 114 has requested acquisition of execution result data (step A4). When receiving a request to acquire execution result data, the job execution control unit 113 changes the allocated resources of the job execution system 150 through the virtualization base 160 (step A5). In step A<b>5 , the job execution control unit 113 changes the allocated resource of the job execution system 150 to the combination of the allocation amount of each resource specified by the allocated resource prediction unit 114 .

After changing the allocated resource amount, the job execution control unit 113 inputs the minimum unit of input data and causes the job process 152 to execute processing (step A6). In step A6, the job execution control unit 113 uses the next minimum unit input data acquisition command based on the last processing data, which is included in the input data information of the job definition data, to the job process 152. process the data.

When the execution of the job process 152 is completed, the job execution control section 113 stores the execution result data in the execution result data storage section 115 (step A7). In step A7, job execution control unit 113 associates the combination of allocated resources changed in step A5 with the minimum unit data processing time, and stores them in execution result data storage unit 115 as execution result data. After that, the process returns to step A3, and the job execution control section 113 requests the resource allocation prediction section 114 to predict the resource allocation.

If the job execution control unit 113 determines in step A4 that acquisition of execution result data has not been requested, it changes the allocated resources of the job execution system 150 through the virtualization base 160 (step A8). In step A<b>8 , the job execution control unit 113 changes the allocation resources of the job execution system 150 to the combination of resource allocation amounts returned from the allocation resource prediction unit 114 .

After changing the allocated resource amount, the job execution control unit 113 inputs the remaining data of the input data 151 and causes the job process 152 to execute processing (step A9). In step A9, the job execution control unit 113 causes the job process 152 to process the remaining data using the remaining input data acquisition command based on the last processing data included in the input data information of the job definition data. let it run.

When the execution of the job process 152 is completed, the job execution control section 113 stores the execution result data in the execution result data storage section 115 (step A10). In step A10, job execution control unit 113 associates the combination of allocated resources changed in step A8 with the data processing time, and stores them in execution result data storage unit 115 as execution result data. The above is the overall flow of processing when a job is executed in the job control system.

6 and 7 show the procedure for predicting the allocated resource amount in step A3. The allocation resource prediction unit 114 reads execution result data from the execution result data storage unit 115 based on the ID of the job definition data received from the job execution control unit 113 (step B1 in FIG. 6). However, at this time, if the processing content of the job definition data has changed since the last execution, it is determined that the execution result data cannot be used, and the execution result data is discarded.

The allocated resource prediction unit 114 determines whether or not there is information as the execution result data when the allocated amount of all resources is the maximum value (step B2). At step B2, the allocation resource prediction unit 114 determines whether or not there is execution result data when the CPU is operated at maximum speed (5) and the disk is operated at maximum speed (5), for example. This combination is the combination with the shortest processing time.

If the allocation resource prediction unit 114 determines in step B2 that the information does not exist, the acquisition of the execution result data when the job execution system 150 is allocated a combination in which the allocation amount of each resource is the maximum value is controlled by the job execution control. A request is made to the unit 113 (step B3). In step B3, the allocated resource prediction unit 114 designates a combination of resource allocation amounts in which the CPU and disk are set to the highest speed (5), respectively, and requests the job execution control unit 113 to acquire execution result data. The job execution control unit 113 executes steps A5 to A7 in FIG. 5 and stores the execution result data in the execution result data storage unit 115. FIG.

After obtaining the execution result data, the allocated resource prediction unit 114 determines whether or not the job processing can be completed by the scheduled end time when the allocated amount of all resources is set to the maximum value (step B4). . In step B4, the allocation resource prediction unit 114 calculates the remaining required time from the minimum unit data processing time (fastest processing time) when the allocation amount of all resources is the maximum value and the remaining input data amount, Calculate the remaining time from the current time and scheduled end time.
Remaining required time = Fastest processing time x Remaining amount of input data Remaining time = Scheduled end time - Current time

The allocation resource prediction unit 114 determines whether the job processing can be completed by the scheduled end time based on whether the difference between the remaining time and the remaining required time is 0 or more. If the remaining required time exceeds the remaining time, the allocation resource prediction unit 114 determines that completion by the scheduled end time is impossible. In that case, the allocation resource prediction unit 114 raises an alarm using notification means such as e-mail (step B5). In addition, the resource allocation prediction unit 114 returns to the job execution unit a combination when the allocation amount of all resources is the maximum value (step B6).

When the allocated resource prediction unit 114 determines in step B3 that the job processing can be completed by the scheduled end time, the execution result data storage unit 115 stores information on the minimum allocation amount of all resources. (Step B7). At step B6, the allocation resource prediction unit 114 determines whether or not there is execution result data when the CPU is operated at the lowest speed (1) and the disk is operated at the lowest speed (1), for example. This combination is the combination with the longest processing time.

If the allocation resource prediction unit 114 determines in step B6 that the information does not exist, the allocation resource prediction unit 114 performs job execution control to obtain execution result data when allocating the job execution system 150 with a combination in which the allocation amount of each resource is the minimum value. A request is made to the unit 113 (step B8). In step B8, the resource allocation prediction unit 114 designates a resource allocation combination in which the speed of the CPU and the disk is the lowest (1), and requests the job execution control unit 113 to acquire execution result data. The job execution control unit 113 executes steps A5 to A7 in FIG. 5 and stores the execution result data in the execution result data storage unit 115. FIG.

After acquiring the execution result data, the allocated resource prediction unit 114 determines whether or not the job processing can be completed by the scheduled end time when the allocated amount of all resources is set to the minimum value (step B9). . In step B9, the allocated resource prediction unit 114 calculates the remaining required time from the minimum unit data processing time (latest processing time) and the remaining amount of input data when the allocated amount of all resources is the minimum value. , the remaining time is calculated from the current time and the scheduled end time.
Remaining required time = Latest processing time x Remaining amount of input data Remaining time = Scheduled end time - Current time

The allocation resource prediction unit 114 determines whether the job processing can be completed by the scheduled end time based on whether the difference between the remaining time and the remaining required time is 0 or more. If the remaining required time is less than the remaining time, the allocation resource prediction unit 114 determines that the process can be completed by the scheduled end time. In that case, the combination of resource allocation amounts that allows the processing to be completed by the scheduled time at the minimum cost is the combination when the allocation amounts of all resources are the minimum values. Accordingly, the resource allocation prediction unit 114 returns to the job execution unit a combination when the allocation amount of all resources is the minimum value (step B10).

If the resource allocation prediction unit 114 determines in step B9 that the processing cannot be completed by the scheduled end time, it searches for a combination of resource allocation amounts that can complete the job processing by the scheduled end time (step B11). In step B10, the allocated resource prediction unit 114 uses, for example, binary exploration to check the execution result data with the median value of the allocated amount of all resources, and to complete the job processing by the scheduled end time. Find the lowest cost combination of quotas. If there is no execution result data with the median value of the allocation amount of all resources, the allocation resource prediction unit 114 requests the job execution control unit to acquire the execution result data when the combination is actually allocated. Get execution result data.

FIG. 8 shows an example of the process of calculating combinations of allocation amounts of resources in binary exploration. As shown in FIG. 8, assume that the processing time corresponding to the combination "maximum value" is 10 seconds when the CPU and disk are each set to the maximum speed (5). It is also assumed that the remaining data amount at that time is 33 cases and the remaining time is 1490 seconds. In this case, the remaining required time is calculated as 330 seconds, and the remaining required time is less than the remaining time. Therefore, in step B4, it is judged that the process can be completed by the scheduled completion time when each resource is set to the maximum value.

On the other hand, it is assumed that the processing time corresponding to the combination "lowest value" is 120 seconds when the CPU and disk are each set to the lowest speed (1). It is also assumed that the remaining data amount at that time is 32 cases and the remaining time is 1370 seconds. In this case, the remaining required time is calculated to be 3840 seconds, and the remaining required time is longer than the remaining time. Accordingly, in step B9, it is determined that processing cannot be completed by the scheduled completion time when each resource is set to the lowest value, and binary exploration is performed in step B11.

In step B11, the allocation resource prediction unit 114 first checks whether or not there is execution result data when the CPU and disk setting values are set to normal (3), which is the median value. If there is no execution result data when the CPU and disk setting values are set to normal (3), the allocation resource prediction unit 114 performs the execution when the CPU and disk setting values are set to normal (3). The job execution control unit 113 is requested to acquire the result data. The job execution control unit 113 sets the CPU and disk setting values to normal (3), and executes the process for the minimum unit data, thereby storing the execution result data in the execution result data storage unit 115 .

As shown in FIG. 8, assume that the processing time corresponding to the combination "median" is 30 seconds when the CPU and disk are each set to normal (3). It is also assumed that the remaining data amount at that time is 31 cases and the remaining time is 1340 seconds. In this case, the remaining required time is calculated to be 930 seconds, and the remaining required time is less than the remaining time. Accordingly, the allocation resource prediction unit 114 determines that the processing can be completed by the scheduled end time with the combination "median".

If the allocation resource prediction unit 114 determines that processing will be completed within the remaining time in the case of the combination "median value", then, as the median value between the "minimum value" and the "median value", the CPU and disk Check if there is execution result data for low speed (2). As shown in FIG. 8, it is assumed that the processing time is 50 seconds when the CPU and disk are each set to low speed (2), corresponding to the combination "median of lowest value and median". It is also assumed that the remaining data amount at that time is 30 cases and the remaining time is 1290 seconds. In this case, the remaining required time is calculated to be 1500 seconds, and the remaining required time is longer than the remaining time. Therefore, the allocation resource prediction unit 114 determines that the processing cannot be completed by the scheduled end time with the combination "the median value between the lowest value and the median value".

Referring to FIG. 3, there is no median value between the normal (3) CPU and disk combination and the slow (2) CPU and disk combination for either CPU or disk resources. Therefore, the allocation resource prediction unit 114 ends the binary exploration, and in step B11, the combination of normal (3) for the CPU and the disk is the combination of resources that can complete the processing by the scheduled end time, and the cost is the lowest. It is judged to be a low combination of

Returning to FIG. 6, the allocation resource prediction unit 114 finds the lowest-cost combination of allocation amounts that can be processed until the scheduled end time using binary exploration, and then records that combination as the combination with the lowest cost. (Step B12). In step B12, if a combination of normal (3) CPU and disk is found, the allocation resource prediction unit 114 regards that combination as the current minimum cost combination (temporary minimum cost combination). ).

The allocation resource prediction unit 114 determines, as a combination to be confirmed, a combination obtained by lowering the allocation amount of each resource by one level from the minimum cost combination (step B13 in FIG. 7). The allocation resource prediction unit 114 determines whether the execution result data for the combination determined in step B13 exists in the execution result data storage unit 115 (step B14). In step B13, the allocation resource prediction unit 114 confirms, for example, a combination of low speed CPU (2) and normal disk (3) and a combination of normal CPU (3) and low speed disk (2). Determined as a combination of targets. Also, in step B14, allocation resource prediction unit 114 determines whether or not there is execution result data for these combinations.

When the allocation resource prediction unit 114 determines that the execution result data does not exist, it determines whether or not the execution result data can be obtained (step B15). In step B15, the allocation resource prediction unit 114 calculates the remaining required time for data acquisition (remaining required data acquisition time) and the remaining time using, for example, the following formulas, and compares the required remaining time with the remaining time.
Remaining required time=minimum cost processing time×(remaining input data−1)+latest processing time Remaining time=scheduled end time−current time Determine that data acquisition is not possible. If the remaining required time is within the remaining time, the allocation resource prediction unit 114 determines that it is possible to acquire the execution result data.

For example, as shown in FIG. 8, it is assumed that the remaining data amount is 30 cases and the remaining time is 1290 seconds at the time when the combination of allocation amounts of resources with the lowest cost is found by binary search. Also, assume that the current minimum cost combination is a combination of normal (3) CPUs and normal (3) disks. At this time, the minimum cost processing time is 30 seconds, and the processing time (slowest processing time) for the combination "lowest value" is 120 seconds. Therefore, the remaining required time is calculated as 30*(30-1)+120=990. In this case, the remaining time is longer than the required time, so the allocation resource prediction unit 114 determines that execution result data can be obtained.

If the allocation resource prediction unit 114 determines that execution result data can be acquired, it requests the job execution control unit 113 to acquire execution result data for a combination of resource allocation amounts for which execution result data has not been acquired ( Step B16). When the allocation resource prediction unit 114 determines that acquisition of the execution result data is not possible, it does not issue an acquisition request for the execution result data, and regards the recorded current minimum cost combination as the minimum cost combination. In this case, the resource allocation prediction unit 114 returns to the job execution control unit 113 a combination of resource allocation amounts that is regarded as the minimum cost to the job execution control unit 113 (step B22).

If the allocation resource prediction unit 114 determines that the execution result data exists in step B14, or after acquiring the execution result data, the allocation resource prediction unit 114 determines that the processing will be completed within the remaining time in the combination of the allocation amount of each resource to be checked. It is determined whether or not a combination exists (step B17). If there is a combination that will be processed within the remaining time, the allocation resource prediction unit 114 determines whether the combination has a lower cost than the current minimum cost combination (step B18). If the allocation resource prediction unit 114 determines that the cost is low in step B18, it updates the current minimum cost combination (step B19). If the allocation resource prediction unit 114 determines in step B18 that the cost is not low, it does not update the current minimum cost combination.

FIG. 9 shows the process of calculating the minimum-cost allocation amount combination of each resource. For example, the allocation resource prediction unit 114 predicts a combination of a slow CPU (2) and a normal disk (3) and a combination of a normal CPU (3) and a slow disk (2) within a time period. It is determined whether or not the processing can be completed in time. Referring to FIG. 9, the remaining required time is less than the remaining time for both combinations. Therefore, in step B17, allocation resource prediction section 114 determines that there is a combination that allows processing to be completed.

The allocation resource prediction unit 114 compares the cost of each combination with the cost of the combination in which the CPU and the disk recorded as the current minimum cost in step B12 are normal (3) respectively. Referring to FIG. 8, the cost of the combination of normal (3) CPU and disk is "60". Also, referring to FIG. 8, the cost of the combination of the CPU being normal (3) and the disk being slow (2) is "55", and the cost of the combination of the CPU being slow (2) and the disk being normal (3) The cost of the combination is "40". In this case, in step B19, the allocation resource prediction unit 114 updates the current minimum cost combination with the combination of low speed CPU (2) and normal disk (3).

The allocation resource prediction unit 114 newly determines a combination of targets for confirmation of the execution result data based on the combination of all confirmed resource allocation amounts (step B20). In step B20, the allocation resource prediction unit 114 confirms the execution result data for the combinations in which the allocation amount of each resource is raised by one level and the combination in which the allocation amount of each resource is lowered by one level from all the confirmed combinations of the allocation amount of each resource. determine if it is necessary. The allocation resource prediction unit 114 determines a combination whose execution result data needs to be confirmed, and a combination whose execution result data is to be confirmed.

The allocation resource prediction unit 114 determines whether or not all combinations have been confirmed (step B21). Allocation resource prediction unit 114 determines that confirmation of the execution result data is necessary in step B20, and when combinations for which execution result data is to be confirmed are determined, it determines that confirmation of all combinations has not been completed. If the allocation resource prediction unit 114 determines that confirmation of all combinations has not been completed, it returns to step B14 and determines whether execution result data exists.

After that, the allocation resource prediction unit 114 goes through steps B15 to B17, and in step B18, there is a combination that can finish the processing in time and whose cost is lower than the current minimum cost combination in the combination to be confirmed. or not. If the allocation resource prediction unit 114 can complete the processing within the time and there is a combination whose cost is lower than the current minimum cost combination, the current minimum cost combination is updated in step B19. In step B20, the allocation resource prediction unit 114 newly determines a combination of objects whose execution result data is to be confirmed. The allocation resource prediction unit 114 repeatedly executes steps B14 to B21 to search for a minimum cost combination that allows processing to be completed within the time limit. If the allocation resource prediction unit 114 determines in step B21 that all combinations have been confirmed, the process proceeds to step B22 and returns the combination recorded as the current minimum cost to the job execution control unit 113. FIG.

FIGS. 10(a)-(d) show the process of searching for the combination of resource allocations with the lowest cost. In FIGS. 10A to 10D, (X, Y) indicates that X is the CPU setting value and Y is the disk setting value. Also, "●" indicates a combination (first type combination) that can finish the processing within the remaining time and has the lowest cost. "▴" indicates a combination (second type combination) that can complete the process within the remaining time but does not have the lowest cost. “X” indicates a combination (third type combination) that cannot be processed within the remaining time.

FIG. 10(a) shows the search for the combination of resource allocations with the lowest cost in the first loop. As explained above, the combination of a slow CPU (2) and a normal disk (3) and a combination of a normal CPU (3) and a slow disk (2) can process in time. It is determined that it can be finished (see Figure 9). Also, the cost of the combination of slow (2) CPU and normal (3) disk is lower than the cost of the combination of normal (3) CPU and slow (2) disk. Accordingly, the allocation resource prediction unit 114 records (2, 3) as the current minimum cost combination in step B19 of the first loop.

FIG. 10(b) shows combinations in which the allocation amount of each resource is raised or lowered by one step from (3, 2) and (2, 3). In step B20, the allocated resource prediction unit 114 derives a combination in which the allocated amount of each resource is increased or decreased by one step from the combinations (3, 2) and (2, 3) that can finish the processing within the time. The allocation resource prediction unit 114 obtains (3, 1), (2, 2), (1, 3) as a combination of (3, 2) and (2, 3) with the allocation amount of each resource raised or lowered by one step. , (4,2), and (2,4). Since (3, 3) has already been confirmed, it is excluded from the combinations to be derived.

Here, the combination obtained by raising the allocation amount of each resource by one level from the minimum cost combination (first type combination) or a higher cost combination (second type combination) is higher than the minimum cost It is clear that the cost is high. Therefore, the allocation resource prediction unit 114 excludes such combinations from the combinations to be confirmed.

FIG. 10(c) shows the combination of confirmation targets determined in step B20. For example, the cost of the combination (3,2) is higher than the cost of the lowest cost combination (2,3). Accordingly, the allocation resource prediction unit 114 excludes (4, 2), which is a combination in which the resource allocation amount is raised by one step from (3, 2), from the combinations to be confirmed. The allocated resource prediction unit 114 also excludes (2, 4), which is a combination obtained by raising the resource allocation amount by one level from the combination of (2, 3), which is the minimum cost combination, from the combinations to be checked. . As a result, in the first step B20, the allocation resource prediction unit 114 determines each combination of (3, 1), (2, 2), and (1, 3) as the target ear combination to be confirmed.

FIG. 10(d) shows the search for the lowest-cost combination of resource budgets in the second loop. In the second loop, the allocation resource prediction unit 114 calculates the remaining required time for each combination of (3, 1), (2, 2), and (1, 3), and determines whether or not the processing will end within the time. find out if The allocation resource prediction unit 114 determines that processing will not be completed within the time for the combination of (2,2) and (1,3), and that processing will be completed within the time for the combination (3,1). Suppose In that case, the allocation resource prediction unit 114 compares the cost of (3, 1) with the cost of (2, 3), which is the combination of the current minimum costs, in step B18. In this case, the cost of (3,1) is higher than the cost of (2,3), so the current minimum cost combination is not updated.

The allocation resource prediction unit 114 determines a combination to be checked from each combination of (3,1), (2,2), and (1,3) in step B20 of the second loop. At this time, if there is a combination (a combination of the third type) that cannot be processed within the time limit among the combinations to be the generation sources, a combination with the allocated amount of resources lowered by one level from that combination can be processed within the time limit. It's clear that it's not over. Therefore, the resource allocation prediction unit 114 excludes from the combinations to be checked a combination in which the resource allocation amount is one step lower than the combination whose processing cannot be completed within the time limit. In the example of FIG. 10(d), the combination of (2, 2) and (1, 3) does not finish the processing within the time, so the combination with the resource allocation amount lowered by one level from these combinations is confirmed. excluded from the target combination.

After that, the allocated resource prediction unit 114 determines whether it is necessary to confirm the execution result data for the combination with the resource allocation amount raised by one level and the combination with the resource allocation amount lowered by one level from all the confirmed resource allocation amount combinations. Make a decision and check the execution result data. If there is a combination whose processing is completed within the time and whose cost is lower than the cost of the current minimum cost combination, the allocation resource prediction unit 114 records that combination as the current minimum cost combination. When there are no more combinations whose execution result data are to be checked, the allocation resource prediction unit 114 determines in step B21 that all combinations have been checked. In this case, the allocation resource prediction unit 114 returns to the job execution control unit 113 in step B22 the minimum cost combination recorded at the time when all combinations have been confirmed.

In this embodiment, the job execution control unit 113 associates the processing time during job execution with a combination of resource allocation amounts of the job execution system 150, and stores them as execution result data in the execution result data storage unit 115. FIG. The allocation resource prediction unit 114 uses the execution result data to predict a combination of resource allocation amounts that will allow processing of the input data 151 to be completed by the scheduled end time at the minimum cost. The job execution control unit 113 controls the allocation resource amount of the job execution system 150 to a combination of resource allocation amounts predicted by the allocation resource prediction unit 114 . By doing so, when performing a job such as batch processing in which the data amount of the input data 151 changes daily using cloud computing, the job can be executed with a combination of the minimum cost and the estimated resource allocation amount. can be implemented.

In this embodiment, the allocation resource prediction unit 114, when the execution result data used for prediction of the resource allocation combination that minimizes the cost is not stored in the execution result data storage unit 115, acquires the execution result data. A request is made to the job execution control unit 113 . The job execution control unit 113 causes the job process 152 to process minimum unit data, measures the processing time, and stores the execution result data in the execution result data storage unit 115 . By doing so, the allocation resource prediction unit 114 can refer to the execution result data required for prediction of the resource allocation amount combination that minimizes the cost. After a certain amount of execution result data has been collected in the execution result data storage unit 115, the allocation resource prediction unit 114 can predict a resource allocation combination that minimizes the cost at the start of job execution.

Note that in the above embodiment, CPUs and disks are considered as resources whose allocation amounts can be changed, but the present disclosure is not limited to this. A resource whose allocation amount is predicted by the allocation resource prediction unit 114 may be a resource that affects the processing time of batch processing and may be another resource that can be changed using the virtualization infrastructure 160 .

Further, in the above embodiment, the allocation resource prediction unit 114 has described an example in which the resource allocation amount combination that minimizes the cost is predicted by comparing the "costs", but the present disclosure is limited to this. not. The allocated resource prediction unit 114 may use some index value related to cost to predict a combination of resource allocation amounts that minimizes the cost. For example, an index value that increases as the cost decreases can be used as the index value. In this case, the resource allocation prediction unit 114 may predict a combination of resource allocation amounts that maximizes the evaluation value.

In the above embodiments, the job control system 110 can be configured as a computer device. FIG. 11 shows a configuration example of an information processing device (computer device) that can be used in the job control system 110. As shown in FIG. The information processing apparatus 500 has a control unit (CPU) 510 , a storage unit 520 , a ROM (Read Only Memory) 530 , a RAM (Random Access Memory) 540 , a communication interface (IF) 550 and a user interface 560 .

The communication interface 550 is an interface for connecting the information processing apparatus 500 and a communication network via wired communication means or wireless communication means. The information processing device 500 can access the virtualization infrastructure 200 (see FIG. 2) via the communication interface 550 . User interface 560 includes a display, such as a display. The user interface 560 also includes input units such as a keyboard, mouse, and touch panel.

The storage unit 520 is an auxiliary storage device capable of holding various data. The storage unit 520 is not necessarily a part of the information processing device 500, and may be an external storage device or a cloud storage connected to the information processing device 500 via a network. ROM 530 is a non-volatile storage device. For the ROM 530, for example, a semiconductor storage device such as a flash memory having a relatively small capacity is used. Programs executed by the CPU 510 may be stored in the storage unit 520 or the ROM 530 .

The above program can be stored using various types of non-transitory computer-readable media and supplied to the information processing apparatus 500 . Non-transitory computer-readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media such as flexible disks, magnetic tapes, or hard disks, magneto-optical recording media such as magneto-optical discs, compact discs (CDs), or digital versatile disks (DVDs). and semiconductor memory such as mask ROM, PROM (programmable ROM), EPROM (erasable PROM), flash ROM, or RAM. The program may also be delivered to the computer using various types of transitory computer readable media. Examples of transitory computer-readable media include electrical signals, optical signals, and electromagnetic waves. Transitory computer-readable media can deliver the program to the computer through wired channels, such as wires and optical fibers, or wireless channels.

RAM 540 is a volatile storage device. Various semiconductor memory devices such as DRAM (Dynamic Random Access Memory) or SRAM (Static Random Access Memory) are used for the RAM 540 . RAM 540 can be used as an internal buffer that temporarily stores data and the like. The CPU 510 expands a program stored in the storage unit 520 or the ROM 530 to the RAM 540 and executes it. The CPU 510 executes programs to implement functions such as the input data confirmation unit 112, the job execution control unit 113, and the allocation resource prediction unit 114, for example.

Although the embodiments of the present disclosure have been described in detail above, the present disclosure is not limited to the above-described embodiments, and changes and modifications can be made to the above-described embodiments without departing from the scope of the present disclosure. are also included in the present disclosure.

For example, some or all of the above-described embodiments may be described in the following supplementary remarks, but are not limited to the following.

[Appendix 1]
Job execution control means for causing a virtual machine configured to be able to change the allocation amount of two or more resources to execute a job for input data;
completing the job by the scheduled end time based on the execution result data including the combination of the resource allocation amount and the processing time of the input data in a predetermined unit, the data amount of the input data, and the scheduled end time; and resource allocation prediction means for predicting a combination of resource allocations that minimizes a cost determined based on the combination of resource allocations,
A job control system in which the job execution control means controls the resource allocation amount of the virtual machine to a combination of the predicted resource allocation amounts.

[Appendix 2]
The job execution control means measures a processing time for executing the job, associates the combination of the resource allocation amount with the measured processing time, and stores the result as the execution result data in the execution result data storage means;
Supplementary Note 1: The resource allocation predicting means searches for a combination of the resource allocation amounts for which the job ends by the scheduled end time, using the execution result data stored in the execution result data storage means. job control system according to .

[Appendix 3]
The allocation resource prediction means reads the execution result data from the execution result data storage means, and based on the processing time included in the read execution result data and the amount of remaining input data, the resources of the virtual machine. calculating the remaining required time of the job when the allocation amount is a combination of the resource allocation amounts of the execution result data, and comparing the remaining required time with the remaining time determined based on the scheduled end time; The job control system according to appendix 2, wherein it is determined whether or not the job will be finished by the scheduled finish time.

[Appendix 4]
The resource allocation prediction means reads execution result data corresponding to a combination that minimizes the allocation amount of each resource from the execution result data storage means, and processes time and remaining input data included in the read execution result data. 4. The job control system according to appendix 2 or 3, which determines whether or not the job will end by the scheduled end time when the resource allocation amount of the virtual machine is set to the combination based on the amount of .

[Appendix 5]
If the execution result data storage means does not store the execution result data corresponding to the combination with the minimum allocation amount of each resource, the allocation resource prediction means causes the job execution control means to obtain the execution result data. ask for
When requested to acquire the execution result data, the job execution control means controls the allocation amount of the resources of the virtual machine to a combination that minimizes the allocation amount of each resource, and gives the virtual machine the predetermined 5. The job control system according to appendix 4, wherein a job is executed for unit input data, a processing time is measured, the execution result data is generated, and the execution result data storage means stores the job.

[Appendix 6]
When the allocation resource prediction means determines that the job will be completed by the scheduled end time when the allocation amount of the resources of the virtual machine is a combination that minimizes the allocation amount of each resource, the combination is: 6. The job control system according to appendix 4 or 5, wherein the allocation amount of the resources is combined to minimize the cost.

[Appendix 7]
The resource allocation prediction means performs binary exploration when it is determined that the job will not end by the scheduled end time when the resource allocation amount of the virtual machine is a combination that minimizes the allocation amount of each resource. 7. The method according to any one of appendices 4 to 6, wherein a combination of the resource allocation amounts for which processing is completed within a time limit is found, and the discovered combination of resource allocation amounts is recorded as a provisional minimum cost combination. Job control system.

[Appendix 8]
The allocated resource prediction means determines a combination of one or more resource allocation amounts as a combination to be confirmed based on the combination of resources recorded as the provisional minimum cost combination, Determining whether or not processing will be completed by the scheduled end time for each of the combinations, and if the combination to be confirmed includes a combination of resource allocation amounts for which processing will be completed by the scheduled end time, the scheduled end time 8. The job control system according to appendix 7, wherein a combination with the lowest cost is specified from among combinations of resource allocation amounts for which processing is completed by, and the specified combination of resources is recorded as the temporary minimum cost combination.

[Appendix 9]
The allocation resource prediction means repeatedly determines the combination to be confirmed and determines whether or not the processing for each of the combinations to be confirmed is completed by the scheduled end time, and determines the combination to be confirmed for the first time. In the determination of, the combination of confirmation targets is determined based on the combination of the resource allocation amounts recorded as the provisional minimum cost combination, and in the second and subsequent determination of the combination of confirmation targets, the previous confirmation target 8. The job control system according to appendix 8, wherein the combination to be confirmed is determined based on the combination of .

[Appendix 10]
In the determination of the combination to be confirmed for the first time, the allocation resource prediction means makes the allocation amount of each resource one step lower than the combination of the allocation amount of the resources recorded as the combination of the provisional minimum cost. One or more combinations are determined as the combinations to be confirmed, and in the second and subsequent determinations of the combinations to be confirmed, the allocation amount of each resource is one step lower than the previous combination to be confirmed. 10. The job control system according to appendix 9, wherein one or more combinations of an allocation amount and an allocation amount one level higher are determined as the combination to be confirmed.

[Appendix 11]
The allocation resource prediction means determines whether or not the processing will be completed by the scheduled end time, in which each of the combinations to be confirmed will be processed by the scheduled end time and the provisional minimum cost will be reduced. A first type combination identified as a combination, a second type combination whose processing is completed by the scheduled end time and the cost is greater than the combination identified as the provisional minimum cost combination, or It is identified which of the combinations of the third type for which the processing is not completed by the scheduled end time, and in the determination of the combination to be confirmed for the second and subsequent times, the allocation amount of each resource is determined according to the first type and the combination of the allocation amount one step higher than the combination of the second type, and the combination of the allocation amount of each resource being one step lower than the combination of the third type, the confirmation target 11. The job control system according to appendix 10, which is excluded from the combination.

[Appendix 12]
The allocation resource prediction means determines whether or not the processing for each of the confirmation target combinations is completed by the scheduled end time until the confirmation target combinations are exhausted. Supplementary note 9: when there is no combination to be confirmed, the combination of resource allocation amounts recorded as the provisional minimum cost combination is set as the combination of resource allocation amounts that minimizes the cost. 12. The job control system according to any one of 11 to 11.

[Appendix 13]
The allocation resource prediction means determines whether execution result data corresponding to the combination to be confirmed is stored in the execution result data storage means. to obtain the execution result data,
When requested to acquire the execution result data, the job execution control means causes the virtual machine to execute the job for the previous predetermined unit of input data, measures the processing time, generates the execution result data, and executes the job. 13. The job control system according to any one of appendices 8 to 12, which is stored in the result data storage means.

[Appendix 14]
When determining that the execution result data storage means does not store the execution result data corresponding to the combination to be confirmed, the allocation resource prediction means determines whether or not the execution result data can be obtained, requesting the job execution control means to obtain the execution result data if it is determined that the execution result data can be obtained; and if determining that the execution result data cannot be obtained, the provisional minimum cost 14. The job control system according to appendix 13, wherein the combination of the resource allocation amounts recorded as the combination of .

[Appendix 15]
Execution result data including a combination of the resource allocation amounts of two or more resource allocation amounts of a virtual machine configured to be changeable and a processing time for input data in a predetermined unit; Allocating the resources so that the job will be completed by the scheduled end time based on the data amount of the target input data and the scheduled end time, and the cost determined based on the combination of the resource allocation amounts will be the minimum. Predict quantity combinations,
controlling the resource allocation of the virtual machine to a combination of the predicted resource allocation;
A job control method for causing the virtual machine to execute a job for the input data.

[Appendix 16]
Execution result data including a combination of the resource allocation amounts of two or more resource allocation amounts of a virtual machine configured to be changeable and a processing time for input data in a predetermined unit; Allocating the resources so that the job will be completed by the scheduled end time based on the data amount of the target input data and the scheduled end time, and the cost determined based on the combination of the resource allocation amounts will be the minimum. Predict quantity combinations,
controlling the resource allocation of the virtual machine to a combination of the predicted resource allocation;
A program for causing a computer to execute processing for causing the virtual machine to execute a job for the input data.

10: Job control system 11: Job execution control unit 12: Allocation resource prediction unit 15: Execution result data 20: Virtual machine 100: System 110: Job control system 111: Job definition data storage unit 112: Input data confirmation unit 113: Job Execution control unit 114: Allocation resource prediction unit 115: Execution result data storage unit 150: Job execution system 151: Input data 152: Job processes 160, 200: Virtualization base 500: Information processing device 510: CPU
520: storage unit 530: ROM
540: RAM
550: Communication interface 560: User interface

Claims (8)

Job execution control means for causing a virtual machine configured to be able to change the allocation amount of two or more resources to execute a job for input data;
The job is completed by the scheduled end time based on the execution result data including the combination of the resource allocation amount and the processing time for the input data in a predetermined unit, the data amount of the input data, and the scheduled end time. allocation resource prediction means for predicting a combination of the resource allocation amounts that minimizes the cost determined based on the combination of the resource allocation amounts;
The job execution control means controls the resource allocation amount of the virtual machine to a combination of the predicted resource allocation amounts ,
moreover,
The job execution control means measures a processing time for executing the job,
The resource allocation prediction means uses the execution result data in which the combination of resource allocation amounts and the measured processing time are associated with each other, for the combination of resource allocation amounts, by the scheduled end time. Look for a combination that ends the job,
The resource allocation prediction means estimates the resource allocation amount of the virtual machine based on the processing time included in the execution result data corresponding to the combination with the minimum allocation amount of each resource and the amount of remaining input data. A job control system that determines whether or not the job will be finished by the scheduled finish time when the combination is set . The resource allocation prediction means performs binary exploration when it is determined that the job will not end by the scheduled end time when the resource allocation amount of the virtual machine is a combination that minimizes the allocation amount of each resource. 2. The job control system according to claim 1 , further comprising: finding a combination of resource allocation amounts for which processing is completed within a time limit, and recording the discovered combination of resource allocation amounts as a provisional minimum cost combination. The allocated resource prediction means determines a combination of one or more resource allocation amounts as a combination to be confirmed based on the combination of resources recorded as the provisional minimum cost combination, Determining whether or not processing will be completed by the scheduled end time for each of the combinations, and if the combination to be confirmed includes a combination of resource allocation amounts for which processing will be completed by the scheduled end time, the scheduled end time 3. The job control system according to claim 2 , wherein a combination with the lowest cost is specified from a combination of resource allocation amounts whose processing is completed by, and the specified combination of resources is recorded as the provisional minimum cost combination. . The allocation resource prediction means repeatedly determines the combination to be confirmed and determines whether or not the processing for each of the combinations to be confirmed is completed by the scheduled end time, and determines the combination to be confirmed for the first time. In the determination of, the combination of confirmation targets is determined based on the combination of the resource allocation amounts recorded as the provisional minimum cost combination, and in the second and subsequent determination of the combination of confirmation targets, the previous confirmation target 4. The job control system according to claim 3 , wherein the combination to be confirmed is determined based on the combination of . In the determination of the combination to be confirmed for the first time, the allocation resource prediction means makes the allocation amount of each resource one step lower than the combination of the allocation amount of the resources recorded as the combination of the provisional minimum cost. One or more combinations are determined as the combinations to be confirmed, and in the second and subsequent determinations of the combinations to be confirmed, the allocation amount of each resource is one step lower than the previous combination to be confirmed. 5. The job control system according to claim 4 , wherein one or more combinations of an allocation amount and an allocation amount one level higher are determined as the combination to be confirmed. The allocation resource prediction means determines whether or not the processing for each of the confirmation target combinations is completed by the scheduled end time until the confirmation target combinations are exhausted. and when there is no combination to be confirmed, the combination of resource allocation amounts recorded as the provisional minimum cost combination is set as the combination of resource allocation amounts that minimizes the cost. 6. The job control system according to 4 or 5 . Execution result data including a combination of resource allocation amounts and a processing time for a predetermined unit of input data in a virtual machine configured such that the allocation amounts of two or more resources can be changed; based on the data amount of the input data to be the target of the job and the scheduled end time, the cost determined based on the combination is the smallest among the combinations of the allocation amounts of the resources that the job will end by the scheduled end time. Predicting a combination of the allocation amounts of the resources that becomes
the computer controls the resource allocation of the virtual machine to a combination of the predicted resource allocation;
the computer causing the virtual machine to execute a job on the input data ;
wherein the computer measures a processing time for executing the job, associates the combination of resource allocation amounts with the measured processing time, and stores the result as the execution result data;
Predicting a combination of resource allocations includes:
The computer uses execution result data in which the combination of resource allocation amounts and the measured processing time are associated to complete the job by the scheduled end time for the combination of resource allocation amounts. looking for a combination; and
The computer determines the resource allocation amount of the virtual machine based on the processing time included in the execution result data corresponding to the combination with the minimum allocation amount of each resource and the amount of remaining input data corresponding to the combination. determining whether or not the job will be finished by the scheduled finish time when the job is finished . Execution result data including a combination of the resource allocation amounts of two or more resource allocation amounts of a virtual machine configured to be changeable and a processing time for input data in a predetermined unit; Based on the data amount of the target input data and the scheduled end time, the cost determined based on the combination is the minimum among the combinations of the allocation amounts of the resources with which the job ends by the scheduled end time. Predict resource allocation combinations,
controlling the resource allocation of the virtual machine to a combination of the predicted resource allocation;
causing the virtual machine to execute a job on the input data ;
measuring the processing time for executing the job;
In predicting a combination of resource allocation amounts,
Searching for a combination of resource allocation amounts in which the job ends by the scheduled end time, using execution result data in which the combination of resource allocation amounts and the measured processing time are associated with each other;
Based on the processing time included in the execution result data corresponding to the combination with the minimum allocation amount of each resource and the amount of remaining input data, when the allocation amount of the resources of the virtual machine is set to the combination, A program for causing a computer to execute processing for determining whether or not the job will be completed by the scheduled end time .

Images