JP2019179280A - Job execution management system and job execution management method - Google Patents

Abstract

To reduce an execution time for an entire job network by avoiding waiting for a lock on exclusive resources as much as possible even when there are a plurality of jobs using the exclusive resources.SOLUTION: A job execution management system includes a manager apparatus for managing a job, and one or more agent apparatuses for executing the job. The manager apparatus identifies and preferentially executes the job that can start using an exclusive resource as soon as the exclusive job is released from among the all jobs queued in a queue of the agent apparatus. The agent apparatus, when the job using the exclusive resource reaches an execution order in the queue without being preferentially executed, holds an execution position of the job to a next execution position if there is an empty in the queue and causes the subsequent job to execute first.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a job execution management system and a job execution management method.

  In the operation management field, a job execution management system that manages job execution is known. A job is a processing procedure defined for a certain purpose such as business. The job is a single or continuous program described by a shell script in UNIX (registered trademark) or a batch script in Windows (registered trademark), for example. The job execution management system includes a job network indicating the execution order of one or more jobs and a program group for processing the batch job.

  In general, a job execution management system is a system in which a manager device executes management of job definition information and job network flow control, and causes an agent device to execute an actual job.

  When executing jobs, resources such as CPU (Central Processing Unit), memory, and disk units are consumed, so the number of jobs that can be executed at the same time is limited and the number of executions is limited so that resources are not burdened. Has been. A queue is often used to limit the number of executions. A queue is a so-called queue, and is a mechanism for controlling an execution wait state that occurs when a job execution input speed and an execution completion speed are different. For example, when a job requested by the manager device is executed on the agent device, if another job is already executed on the agent device, multiple jobs will be executed simultaneously, which will cause a resource shortage. There is a fear. Therefore, the execution number limit prevents the depletion of resources by limiting the number of jobs that can be executed at the same time, waiting in turn in the queue, and executing when the order comes.

  However, there are not only resources that can be shared by a plurality of jobs, such as a CPU and memory, but also resources that cannot be shared, in other words, resources that are occupied by any job (hereinafter referred to as “exclusive resources”). Exclusive resources include, for example, a printer in print processing, a backup destination area or tape device in data backup processing, or transmission of huge data in data transmission processing to a server in another geographically distant data center It is. Regarding the transmission of huge data, the network communication itself can be shared, but the transmission of huge data will substantially occupy the network during the transmission period, so the transmission of huge data is exclusive. It can be regarded as a resource. In addition, depending on the requirements of the application used in the job, there may be a thing that requires the use of exclusive resources.

  Patent Document 1 describes a technique for optimizing execution scheduling based on an allowance time when performing execution order control between tasks. The allowance time means that if task C, whose occupation time length of the shared resource is known, occupies a resource shared with task A whose execution start scheduled time is determined, task A is executed after task C releases the resource. It shows the time margin until it is started.

JP 2003-131892 A

  If jobs that use the same exclusive resource try to run at the same time, a conflict occurs, and one of the jobs must wait. Since such waiting becomes a bottleneck of processing, it leads to an increase in processing time of the entire job network flow. Therefore, the job execution management system requires lock control for allowing the job to lock and use the exclusive resource.

  It is assumed that all lock controls described below are “waiting exclusive locks”. In this method, when the exclusive resource to be locked is already locked by another person, it is not rejected, but is put into a lock order waiting state and locked when the order comes. Further, there are “shared lock” and “exclusive lock” as types of locks, but here “exclusive lock” is targeted. Therefore, “exclusive lock that can be waited for” is targeted.

  Examples of existing lock control include the following first to third methods.

  The first method is, for example, a method in which a lock is acquired when a job execution request is made to the agent device and the job is queued, that is, when the job is not yet executed. In the case of the first method, since the lock can be acquired while waiting for the turn in the queue, the execution of the job can be started immediately using the exclusive resource when the execution order of the job comes around. it can. On the other hand, in the case of the first method, the exclusive resource is locked before the turn comes, so if it takes a long time to execute the job queued in advance, it exists in another queue A job that wants to use the same exclusive resource that can be executed immediately becomes locked.

  In contrast to the first method, the second method waits for the turn in the queue and locks the necessary exclusive resources after the queue is free and the job can be executed. In the second method, the lock waiting state does not occur as in the first method. However, if the lock cannot be acquired immediately and the job is waiting, the job will not start immediately. In addition, since the job is already in the execution state, the limit on the number of executions of the corresponding queue is wasted. In other words, in the case of the second method, there is a job that is waiting for lock but not executed, and therefore, the execution of subsequent jobs in the queue is hindered.

  On the other hand, the third method is a method of locking at the stage where the execution order has come. Specifically, in the third method, when a job that is desired to be locked cannot be locked immediately, the subsequent job is executed first while the lock request is being made, and the “next execution order” This is a method of maintaining the “position where the wheel turns”. In the third method, the above process is repeatedly executed until a lock can be acquired. In the case of the third method, when the lock cannot be performed immediately and the lock is awaited, the execution order is given to the succeeding job, so that the execution of the succeeding job is not hindered. That is, the disadvantages of the second method are eliminated by the third method. However, in the case of the third method, the succeeding job is executed first while acquiring the lock. Therefore, if the job executed first is a job that has been running for a long time, the first job is thereby executed. There is a possibility that the same disadvantage of the method (that is, waiting for lock) may occur.

  As described above, the existing lock control has different timings for locking exclusive resources, but all have some problems, and it is difficult to say that all of them are optimal methods.

  In Patent Document 1, it is assumed that a task is “interruptible”, but there is a case where a job using an exclusive resource cannot be interrupted. Therefore, the technique of Patent Document 1 cannot be applied as it is to a job execution management system that manages jobs as described above.

  The present invention has been made to solve the above-described problem. Even when there are a plurality of jobs that use exclusive resources, the waiting time for locking exclusive resources is avoided as much as possible, and the execution time of the entire job network is further increased. An object is to provide a technique that can be shortened.

  A job execution management system according to the present invention includes a manager device that manages a job and one or more agent devices that execute the job, and the manager device includes all queued queues of the agent device. Among the jobs, the job that can start using the exclusive resource earliest after the exclusive resource lock is released is identified and preferentially executed, and the agent device uses the exclusive resource. When a job reaches the execution order in the queue without being preferentially executed, the execution position of the job is kept at the position where the job is executed next when the queue is empty, and the subsequent job is first Let it run.

  The job execution management method of the present invention is a job execution management system comprising a manager device that manages a job and one or more agent devices that execute the job, and the manager device is queued in the agent device. Among all the queued jobs, the job that can start using the exclusive resource earliest after the lock of the exclusive resource is released is identified and preferentially executed, and the agent device If the job using the exclusive resource reaches the execution order in the queue without being preferentially executed, the execution position of the job is left at the position where the job is executed next when the queue is empty, The subsequent job is executed first.

  According to the present invention, even when there are a plurality of jobs that use exclusive resources, it is possible to avoid waiting for the exclusive resources to be locked as much as possible and to shorten the execution time of the entire job network.

It is a block diagram which shows the structural example of the job execution management system which concerns on the 1st Embodiment of this invention. 3 is a flowchart showing an operation example (job execution management method) of the job execution management system shown in FIG. 1. It is a block diagram which shows the structural example of the job execution management system which concerns on the 2nd Embodiment of this invention. FIG. 4 is an image of first control (priority execution) executed by the job execution management system shown in FIG. 3. It is a flowchart which shows an example of 1st control. It is a flowchart which shows an example of 2nd control. It is a 1st figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 2nd figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 3rd figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 4th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 5th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 6th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 7th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is an 8th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 9th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 10th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is an 11th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 12th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 13th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 14th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 15th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 16th figure which shows the queue state for demonstrating the specific example of the priority control performed in 1st Embodiment. It is a 17th figure which shows the queue state for demonstrating the operation | movement at the time of performing a job with the system (3rd system) of the related technology which does not apply this invention. It is the 18th figure which shows the queue state for demonstrating the operation | movement at the time of performing a job by the system (3rd system) of the related technology which does not apply this invention. It is a 19th figure which shows the queue state for demonstrating operation | movement at the time of performing a job by the system (3rd system) of the related technology which does not apply this invention. FIG. 20 is a twentieth diagram illustrating a queue state for explaining an operation when a job is executed by a related technology method (third method) to which the present invention is not applied. It is a 21st figure which shows the queue state for demonstrating the operation | movement at the time of performing a job by the related technique system (3rd system) to which this invention is not applied. It is a 22nd figure which shows the queue state for demonstrating the operation | movement at the time of performing a job by the related technique system (3rd system) to which this invention is not applied. It is a 23rd figure which shows the queue state for demonstrating the operation | movement at the time of performing a job by the related technique system (3rd system) to which this invention is not applied. It is a 24th figure which shows the queue state for demonstrating the operation | movement at the time of performing a job by the related technique system (3rd system) to which this invention is not applied. It is a 25th figure which shows the queue state for demonstrating the operation | movement at the time of performing a job with the system (3rd system) of the related technology which does not apply this invention. It is the 26th figure which shows the queue state for demonstrating the operation | movement at the time of performing a job by the related technique system (3rd system) to which this invention is not applied.

[First Embodiment]
(Description of configuration)
FIG. 1 is a block diagram showing a configuration example of a job execution management system 100 according to the first embodiment of the present invention.

  The job execution management system 100 includes a manager device 102 that manages jobs and one or more agent devices 104 that execute jobs.

  The manager device 102 identifies and prioritizes a job that can start using the exclusive resource earliest after the lock of the exclusive resource is released from among all jobs queued in the queue of the agent device 104. Let it run.

When the job using the exclusive resource reaches the execution order in the queue without being preferentially executed, the agent device 104 keeps the execution position of the job at the position to be executed next when the queue is empty, First, the subsequent job is executed.
(Description of operation)
FIG. 2 is a flowchart showing an operation example (job execution management method) of the job execution management system 100 shown in FIG.

  The manager device 102 identifies and prioritizes a job that can start using the exclusive resource earliest after the lock of the exclusive resource is released from among all jobs queued in the queue of the agent device 104. This is executed (step S50).

When the job using the exclusive resource reaches the execution order in the queue without being preferentially executed, the agent device 104 keeps the execution position of the job at the position to be executed next when the queue is empty, First, the subsequent job is executed (step S51).
(Explanation of effect)
According to the first embodiment described above, when there are a plurality of jobs using exclusive resources, it is possible to avoid waiting for the exclusive resources to be locked as much as possible and to shorten the execution time of the entire job network. Of course, in that case, there is no need for the user to intentionally adjust the job scheduling time so that the execution timings do not overlap.

The configurations / operations and priority execution of the manager device 102 and the agent device 104 will be described in detail in the second embodiment below.
[Second Embodiment]
(Description of configuration)
FIG. 3 is a block diagram illustrating a configuration example of the job execution management system 1 according to the second exemplary embodiment of the present invention. Note that a general job execution management system generally includes a control unit that activates a job network and controls its flow according to a schedule. However, since such control means is not directly related to the present embodiment, it will be omitted in the following description.

  The job execution management system 1 includes a manager device 2 that mainly executes job definition information management and job network flow control, and an agent device 3 that executes jobs. 3 illustrates the case where the number of agent devices 3 is “3”, but the number of agent devices 3 may be “two or less” or “four or more”. Good.

  The manager device 2 includes a job execution control unit 10, a job definition information DB 11, and an exclusive resource management DB 13. DB is an abbreviation for DataBase.

  The job definition information DB 11 stores, for each job, a job name, a script, an expected execution time that is an expected time from the start of execution of the job to the end of execution, the name of the exclusive resource, or identification information (hereinafter also referred to as an exclusive resource name). Are recorded as job definition information.

  Exclusive resources include, for example, a printer in print processing, a backup destination area or tape device in data backup processing, or transmission of huge data in data transmission processing to a server in another geographically distant data center It is. Of course, the exclusive resource is not limited to the above.

  The exclusive resource management DB 13 displays all the exclusive resource names that can be used in the job execution management system 1, whether or not these exclusive resources are locked, the locked job name if locked, and the lock waiting status. Record the job name.

  The job execution control unit 10 reads job definition information from the job definition information DB 11 and transmits a job execution request (start of queuing) to the agent device 3. The job execution control unit 10 receives a lock acquisition request and a lock release request from the agent device 3 and registers these requests in the exclusive resource management DB 13.

  The job execution control unit 10 has a lock control function for controlling the lock of the exclusive resource represented by the exclusive resource name. The lock control function does not need to be possessed by each application (agent device 3) that handles exclusive resources, but only the manager device 2 may have it. In this case, information “which exclusive resource is used” is defined in advance in the job definition information DB 11 as a parameter. The exclusive resource name may be simple character string information. For example, when the exclusive resource is the printer Z, it is determined in advance that a lock is requested with the character string information “Z”, and the job execution control unit 10 excludes when the lock request “Z” is received. Take control.

  The agent device 3 includes a job execution control unit 20. The job execution control unit 20 performs queue control and execution of a job requested to be executed by the manager device 2. The job execution control unit 20 returns the job execution result to the manager device 2. The job execution control unit 20 transmits a “lock acquisition request” to the manager device 2 when a job using the exclusive resource is executed, and issues a “lock release request” when the execution of the job is completed. Send.

  In the job execution management system 1, priority control is executed in order to shorten a useless lock waiting time of a job that uses exclusive resources. The priority control includes a first control (priority execution) by the job execution control unit 10 and a second control by the job execution control unit 20.

The first control includes the following controls (1) to (4).
(1) Extraction of Jobs that Use Targeted Exclusive Resources The job execution control unit 10 extracts all jobs that use target exclusive resources from among the jobs queued in each agent device 3. When a plurality of such jobs exist in the same queue, the job execution control unit 10 targets only those queued in advance.
(2) Calculation of Remaining Necessary Time The job execution control unit 10 determines the “remaining necessary time” required for the job currently being executed in the queue to end in each queue in which the job of control (1) is queued. Calculate time. The remaining required time is calculated for all target queues.

  The remaining necessary time can be calculated from, for example, an expected execution time of a job being executed and an elapsed time since the execution of the job started. For example, the remaining necessary time can be calculated by the following (Equation 1).

Remaining required time = Expected execution time-Elapsed time (Formula 1)
The job execution time is the time from the start of the job on the agent device 3 to the completion. The expected execution time is an expected value of the execution time. The predicted execution time can be estimated using past results, for example. Since the job execution management system 1 is a system characterized by executing the same processing in the form of daily, weekly, or monthly, it is possible to refer to the previous results. In addition, in the case of a job having a feature that the processing time increases according to the input data amount, the job execution control unit 10 uses the data obtained by formulating the relationship between the input data amount and the processing time. It is also possible to estimate the expected execution time from the input data amount of the job to be executed.

For example, the job execution control unit 10 refers to the job definition information DB 11 and acquires the estimated execution time. Further, the job execution control unit 10 acquires the elapsed time of each job by making an inquiry to each agent device 3, for example.
(3) Determining the queue that can be emptied first The job execution control unit 10 compares the remaining required time of each queue, which is the calculation result of control (2), and determines the queue that can be emptied first. .
(4) Job execution priority change The job execution control unit 10 is queued first among the jobs that use the target exclusive resource and are queued in the queue determined in the control (3). The execution position of the current job is changed to “position to be executed next when the queue is empty” in the corresponding queue. If there is one queued job, the earliest queued job is that one job.

  FIG. 4 is an image of the first control.

  By executing the above controls (1) to (4) each time an exclusive resource becomes available, a job that uses the exclusive resource is executed preferentially and the lock waiting for the exclusive resource is prevented from occurring as much as possible. be able to. The processing of the above controls (1) to (4) is a process of “finding a job that can be started using the exclusive resource earliest when the exclusive resource becomes available and preferentially executing the job” Can be summarized as:

  In addition, when the execution order of the queue is reached without executing the first control for the job using the exclusive resource, the job execution control unit 20 executes the second control.

  In the second control, when a job that uses exclusive resources reaches the execution order in the queue without performing the above-described controls (1) to (4), the execution position of the job is set to “empty in queue”. This is a control that keeps it at the “next execution position if possible” and executes the subsequent job first.

  By the second control, it is possible to prevent the execution of the subsequent job in the corresponding queue from being hindered by the job waiting for the exclusive resource lock.

  However, when executing the second control, there is a possibility that the control (3) and the control (4) will continue to be leaked. In order to avoid this, a correction value is introduced in the calculation of the control (2) that influences the determination of the control (3). This correction value is the excess time (the excess time when the estimated start time from the time when the target job is queued in the agent device 3). For example, when a job is queued, it is possible to calculate how long to wait from the estimated execution time of a previously queued job before the execution of the queued job is started. The excess time is the expected start time starting from the queuing time, and is not executed even if this expected start time is exceeded, i.e., the queue cannot be acquired by priority control of other queues. It is time. Therefore, the job execution control unit 10 does not set the “remaining required time” calculated in the control (2) as a comparison target in the control (3), but subtracts the “excess time” from the “required required time”. The value (that is, the corrected value) is set as a comparison target in the control (3).

  Note that the expected start time can be obtained by, for example, (Equation 2) below.

Estimated start time = Total estimated execution time of queued unexecuted jobs + Expected execution time of currently executing job-Elapsed execution time of the job (Equation 2)
The “excess time” can be obtained by the following (formula 3).

Excess time = elapsed time since queuing-expected start time (Equation 3)
Below, the specific example of 1st control is shown. When the “remaining required time” is calculated as the calculation result of the control (2), it is assumed that the queue A is 20 seconds, the queue B is 30 seconds, and the queue C is 40 seconds. If the correction value is taken into consideration, the queue that can be emptied earliest is queue A. However, the job C0 queued in the queue C reaches the head of the queue without being subject to the control (1) to (4), but the exclusive resource is not available while another job is locked. It is assumed that the subsequent job is executed first by the control (5). When the waiting time (excess time) is, for example, 30 seconds, the “remaining necessary time” after correction of the queue C is 40−30 = 10 seconds. Accordingly, the job to be executed next determined by the control (3) and the control (4) is the job C0 queued in the queue C and waiting for the lock.
(Description of operation)
(Description of general operation)
FIG. 5 is a flowchart illustrating an example of the first control.

  The processing shown in this flowchart is a job execution request to the agent device 3 when the lock of any exclusive resource registered in the exclusive resource management DB 13 is released or for a job for which an exclusive resource name is set. It is executed when.

  The target exclusive resource is an exclusive resource whose lock has been released in the former case, and an exclusive resource set in the job for which execution has been requested in the latter case. In the following description, the exclusive resource name is Z.

  The job execution control unit 10 makes an inquiry to the job execution control unit 20 of each agent device 3, and all the jobs X1 that use the exclusive resource Z among the jobs queued in the queue of each agent device 3. Is extracted (step S1). If there are a plurality of jobs using the exclusive resource Z in the same queue, only the jobs queued in advance are targeted.

  If there is no job X1 (No in step S1), the process of this flowchart ends.

  When one or more jobs X1 exist (Yes in step S1), the job execution control unit 10 calculates the remaining required time X2 of the job currently being executed in each queue to which the job X1 belongs (step S2). The calculation method of the remaining required time X2 is as described above.

  The job execution control unit 10 corrects the remaining necessary time X2, and calculates the remaining necessary time X3 after correction (step S3). Specifically, the job execution control unit 10 calculates an excess time with respect to the expected start time of the job X1. The calculation method of the expected start time and the excess time is as described above. Then, the job execution control unit 10 calculates the corrected remaining necessary time X3 by subtracting the excess time from the remaining necessary time X2.

  The job execution control unit 10 compares the remaining necessary time X3 after correction of each queue, and determines the queue X4 that can be emptied first (step S4). The job execution control unit 10 transmits an instruction to change the job position to the job execution control unit 20 of the agent device 3 having the queue X4.

  Based on the above instruction, the job execution control unit 20 changes the position of the job queued first in the queue X4 among the jobs using the exclusive resource Z to “next executable position” ( Step S5).

  FIG. 6 is a flowchart illustrating an example of the second control.

  The job execution control unit 20 determines whether or not the job using the exclusive resource has reached the execution order in the queue without being preferentially executed (without the first control being executed) (step S10). When the job using the exclusive resource reaches the execution order in the queue without being preferentially executed (Yes in step S10), the job execution control unit 20 sets the execution position of the job to “ And the subsequent job is executed first (step S11).

When the first control is not executed and the job does not reach the queue execution order (No in step S10), and when the process of step S11 is completed, the process of this flowchart ends.
(Description of specific operation)
Hereinafter, priority control will be specifically described with reference to FIGS. 7 to 22 of FIGS. 7 to 32 are diagrams for explaining the present embodiment, and FIGS. 23 to 32 are diagrams for explaining a system of related technology to which the present invention is not applied. FIG.

The premise of the explanation is as follows.
It is assumed that there are three agent devices 3 and each agent device 3 has only one queue (queue A, B, C).
It is assumed that job queuing as shown in FIG. 7 is performed at the start time. FIG. 7 shows that the job is queued in advance as it goes down, and that the job written below the double line is an execution state.
In FIG. 7, bold characters in the job column indicate jobs for which the exclusive resource Z is set (jobs A5, B2, and C5).
In FIG. 7, “Forecast” is a simplified version of the expected execution time.
In FIG. 7, “excess” is a simplified overtime.
In order to make the explanation clearer, it is assumed that the extraction in step S1 is completed, and as a result, the queue A, queue B, and queue C shown in FIG. 7 are always hit.
Assume that job A7 (expected execution time: 20 seconds) is added to queue A after 60 seconds, and job C6 (expected execution time: 20 seconds) is additionally input to queue C after 100 seconds.
{Circle around (3)} Since the exclusive resource Z is empty at the start time (FIG. 7), priority control is executed. The remaining required times as a result of step S2 are A = 10, B = 80, and C = 20, respectively. Since the excess time is all 0, the remaining necessary time is not corrected. Therefore, the queue that can be executed earliest is queue A. Then, the job A5 in the queue A is moved immediately after the job A1 (FIG. 8).
(5) 10 seconds after the start 10 seconds later, the job A1 is completed and the job A5 using the exclusive resource Z is started to be executed (FIG. 9).
(3) 30 seconds after the start 20 seconds after the state shown in FIG. 9, the job A5 is completed and the lock of the exclusive resource Z is released. At this time, priority control is executed again (FIG. 10).

In this case, the comparison targets are job B2 and job C5. Since there is no excess time, the queue that is executed earliest without correction is queue C. Therefore, the job C5 is moved immediately after the job C2 (FIG. 11).
(2) 50 seconds after the start After 20 seconds from the state shown in FIG. 11, the job C2 is completed, and the execution of the job C5 using the exclusive resource Z is started (FIG. 12).
(2) 60 seconds after the start 10 seconds after the state shown in FIG. 12, A7 (expected execution time: 20 seconds) is newly queued in the queue A (FIG. 13).
(3) 80 seconds after start Job B1 is completed 20 seconds after the state shown in FIG. At this time, since the job B2 to be executed next is a job that uses the exclusive resource Z, contention occurs. Therefore, the job execution control unit 20 requests acquisition of lock of the exclusive resource Z (waiting for lock), and starts execution of the subsequent job B3 first (FIG. 14).
(9) 90 seconds after the start 10 seconds after the state shown in FIG. 14, the job C5 is completed and the lock of the exclusive resource Z is released. At this time, priority control is executed again (FIG. 15).

In this case, the comparison targets are job A7 and job B2. Job B2 has an excess time. In consideration of this, the remaining necessary time after correction is queue A = 0 and queue B = 30−10 = 20, respectively, and queue A is smaller. Therefore, the position of job A7 is moved. In this case, since there is no job being executed, execution of job A7 is started as it is (FIG. 16).
(2) 100 seconds after the start 10 seconds after the state shown in FIG. 16, C6 (expected execution time: 20 seconds) is newly queued in the queue C (FIG. 17).
(2) 110 seconds after the start 10 seconds after the state shown in FIG. 17, the job A7 is completed, and the lock of the exclusive resource Z is released. At this time, priority control is executed again (FIG. 18).

In this case, the comparison targets are job B2 and job C6. And B2 has an excess time. Taking this into account, the remaining necessary time after correction is queue B = 10−30 = −20 and queue C = 0, and queue B is smaller. Therefore, the position of job B2 is moved. In this case, since the job B2 is already at the “next execution position”, the movement process is not necessary. Further, since the queue C is not a target for priority control, the execution of C4 is started in order (FIG. 19).
(1) 120 seconds after the start 10 seconds after the state of FIG. 19, the job B3 is completed, and the execution of the job B2 using the exclusive resource Z is started (FIG. 20).
(3) 160 seconds after the start 40 seconds after the state shown in FIG. 20, the job C4 is completed. At this time, since the job C6 to be executed next is a job that uses the exclusive resource Z, contention occurs. Therefore, the job execution control unit 20 requests acquisition of the exclusive resource Z lock. Since there is no job that can be subsequently executed in the queue C, the queue C is a queue in which no job is executed while the job C6 stays at the “position to be executed next” (FIG. 21).
(2) 170 seconds after the start 10 seconds after the state shown in FIG. 21, the job B2 is completed and the lock of the exclusive resource Z is released. Since the exclusive resource Z is already waiting for the lock of the job C6, the job C6 acquires the lock and shifts to the execution state (FIG. 22).
(2) After 190 seconds from the start After 20 seconds from the state shown in FIG. 22, the job C6 is completed and all jobs are completed.

  As described above, in the present embodiment, all jobs are completed in a total of 190 seconds.

Next, in order to show the effect of the present embodiment, an operation and a total execution time when a job is executed by a related art method to which the present invention is not applied will be described. As this method, the third method (see [Problems to be Solved by the Invention]), which has the least demerit, is adopted. The preconditions are the same as those described above. In addition, a “lock” column is described instead of the “excess” column, and the waiting order for acquiring the lock of the exclusive resource Z is described here. It is indicated that the job whose value is 1 is currently locking the exclusive resource Z.
(3) Start time The start time is shown in FIG. 7, and the processes are executed in this order from the queued state.
(2) 60 seconds after the start 60 seconds after the state shown in FIG. 7, a new job A7 (expected execution time: 20 seconds) is queued in the queue A (FIG. 23).
(2) 70 seconds after the start 10 seconds after the state shown in FIG. 23, the job A4 is completed. Subsequent job A5 requests a lock because it is a job that requires exclusive resource Z, but since there is no other job that locks exclusive resource Z, the lock can be acquired and execution is started as it is (FIG. 24). .
(3) 80 seconds after start Job B1 is completed 10 seconds after the state shown in FIG. At this time, since the job B2 to be executed next is a job that uses the exclusive resource Z, contention occurs. Therefore, the lock acquisition request for the exclusive resource Z is executed, and the subsequent job B3 is started first (FIG. 25).
(2) 90 seconds after the start 10 seconds after the state shown in FIG. 25, the job A5 is completed and the lock of the exclusive resource Z is released. Since the exclusive resource Z has already requested the lock by the job B2, the job B2 acquires the lock of the exclusive resource Z (FIG. 26).
(1) 100 seconds after the start 10 seconds after the state shown in FIG. 26, C6 (expected execution time: 20 seconds) is newly queued in the queue C (FIG. 27).
(2) 110 seconds after the start 10 seconds after the state shown in FIG. 27, the job A6 is completed. Since the succeeding job A7 is a job that uses the exclusive resource Z, a lock request is made. However, since the job B2 has already acquired the lock, a waiting occurs. In queue A, there is no job that can be executed subsequently. Therefore, the queue A is a queue that does not execute any job (FIG. 28).
(1) 120 seconds after the start 10 seconds after the state shown in FIG. 28, the job B3 and the job C4 are completed. The succeeding job B2 is a job that uses the exclusive resource Z, but since the lock has already been acquired, the execution is started as it is. The subsequent job C5 is also a job that uses the exclusive resource Z and requests a lock, but a waiting occurs because another job is already locked. Since job C6 subsequent to job C5 is also a job that uses exclusive resource Z and cannot be executed in advance, queue C remains as it is, and nothing can be executed (FIG. 29).
(1) 170 seconds after the start 50 seconds after the state shown in FIG. 29, the job B2 is completed and the lock of the exclusive resource Z is released. Subsequently, the job A7 acquires the lock of the exclusive resource Z and starts to be executed (FIG. 30).
(2) 190 seconds after the start After 20 seconds from the state shown in FIG. 30, the job A7 is completed and the lock of the exclusive resource Z is released. Subsequently, the job C5 acquires the lock of the exclusive resource Z and starts to be executed (FIG. 31).
(3) 230 seconds after the start 40 seconds after the state shown in FIG. 31, the job C5 is completed and the lock of the exclusive resource Z is released. Subsequently, the job C6 acquires the lock of the exclusive resource Z and starts executing (FIG. 32).
(2) After 250 seconds from the start After 20 seconds from the state of FIG. 32, the job C6 is completed and the execution of all jobs is completed.
(Explanation of effect)
As described above, in the third method, execution of all jobs is completed in a total of 250 seconds.

  When this embodiment is compared with the third method, there is a difference of 250-190 = 60 seconds. In other words, according to this embodiment, the execution time of the entire job is improved by about 25% compared to the third method, which is considered to have the least disadvantage among the methods used before this embodiment. I understand that. In other words, according to the second embodiment described above, when there are a plurality of jobs that use exclusive resources, it is possible to avoid waiting for the exclusive resources to be locked as much as possible and to shorten the execution time of the entire job network. . Of course, in that case, there is no need for the user to intentionally adjust the job scheduling time so that the execution timings do not overlap.

  As mentioned above, although this invention was demonstrated using each embodiment, the technical scope of this invention is not limited to description of said each embodiment. It is obvious to those skilled in the art that various modifications or improvements can be added to the above embodiments. Therefore, it is needless to say that embodiments with such changes or improvements are also included in the technical scope of the present invention. The numerical values and names of the components used in the embodiments described above are illustrative and can be changed as appropriate.

DESCRIPTION OF SYMBOLS 1 Job execution management system 2 Manager apparatus 3 Agent apparatus 10 Job execution control part 11 Job definition information DB
13 Exclusive resource management DB
20 Job Execution Control Unit 100 Job Execution Management System 102 Manager Device 104 Agent Device

Claims (5)

A manager device for managing jobs;
One or more agent devices that execute the job,
The manager device identifies, from among all the jobs queued in the queue of the agent device, the job that can start using the exclusive resource as soon as the exclusive resource lock is released. To give priority execution,
When the job using the exclusive resource reaches the execution order in the queue without being preferentially executed, the agent device executes the job execution position next when the queue is free. The job execution management system is characterized in that the job following the job is executed first. The manager device is
Extracting all the jobs that use the exclusive resource among the jobs queued in the respective queues of all the agent devices,
By calculating the remaining required time of the job currently being executed in the queue to which the extracted job belongs, and subtracting the excess time, which is the time waiting for the lock from being acquired, from the remaining required time Calculate the remaining required time after correction,
Compare the remaining necessary time after correction of each of the queues, determine the queue that can be emptied first,
The execution position of the job that is queued first among the jobs that use the exclusive resource in the queue for the agent device that has the queue that can be freed first is free in the queue. The job execution management system according to claim 1, further comprising: instructing to change to a position to be executed next.   The remaining required time is a time obtained by subtracting an elapsed time from the start of execution of the job from an expected execution time that is an expected value of the time from the start of the job to the completion on the agent device. The job execution management system according to claim 2, wherein:   4. The job execution management system according to claim 2, wherein the excess time is a time obtained by subtracting an expected start time from an elapsed time after the job is queued. A job execution management system comprising: a manager device that manages a job; and one or more agent devices that execute the job,
The manager device identifies, from among all the jobs queued in the queue of the agent device, the job that can start using the exclusive resource as soon as the exclusive resource lock is released. To give priority execution,
When the job using the exclusive resource reaches the execution order in the queue without being preferentially executed, the agent device executes the job execution position next when the queue is free. A job execution management method characterized in that the job is retained at a position where the job is executed and the subsequent job is executed first.

Images