JP5958111B2 - Job management program, job management method, and job management apparatus - Google Patents

Description

  This case relates to a job management program, a job management method, and a job management apparatus.

  Conventionally, there is a job management apparatus for executing a plurality of computers (calculation nodes) in parallel. When a job is input by a user, the job management apparatus assigns the job to one of the computation nodes (scheduling). Then, the job management apparatus transfers the job and data used in the job to the storage area of the calculation node that executes the job processing, so that the job is executed in the calculation node, and the job result is the user (terminal). Is output for.

Japanese Patent Laid-Open No. 10-40170

  Among jobs, there are jobs that perform processing on large-scale DBs. At present, since the job management apparatus manages jobs on a job basis, it is necessary to transfer a large-scale DB to the storage area of the computation node for each job. That is, when one job and another job are executed on the same calculation node, the job management apparatus once deletes the large-scale DB from the storage area of the calculation node when one job is completed, When executing, the large-scale DB used for the other job is stored again in the storage area of the computation node. In this case, since the large-scale DB may reach several hundred GB, it takes a lot of time to store the large-scale DB in the storage area of the computation node.

  Note that Patent Document 1 discloses that when data used in the past by a program remains in the cache, the data can be used if another program can use the data. However, since the job management apparatus deletes the data in the storage area of the computation node every time the job ends, it is difficult to apply the technique of Patent Document 1 to the job management apparatus.

  In one aspect, an object of the present invention is to provide a job management program, a job management method, and a job management apparatus capable of performing efficient job processing.

Job management programs described herein, the content of the job, execution of the said job information of the common data and said common data different from the individual data information corresponding to a plurality of jobs to be utilized in the job, it is included get the instruction, based on the execution command of the job acquired, allocating the job to one of the plurality of computing nodes, individual and common data used in the execution of a plurality of jobs to compute node assigned the job transfers the data to instruct the execution of the job, with respect to computing node assigned the job, instructs the deletion of the individual data of the job execution is completed in the calculation node assigned the job, common data and the next meter assigned the job of the job execution is completed in the calculation node assigned the job If the common data of the job that will be executed by the node matches, and other jobs that are running a common data of the job execution is completed in the calculation node assigned the job compute nodes assigned the job If not at least for one of the case of using the run, indicating the deletion of the common data of the job for the computing nodes assigned the job, a job management program for executing the processing to the computer.

Job management methods described herein, the content of the job, execution of the said job information of the common data and said common data different from the individual data information corresponding to a plurality of jobs to be utilized in the job, are included an acquisition step of acquiring an instruction, based on the execution command of the job acquired, allocating the job to one of the plurality of computing nodes, used in executing a plurality of jobs to compute node assigned the job common It transfers the data and individual data, and the execution instruction step for instructing the execution of the job, with respect to computing node assigned the job, the individual data of the job execution is completed in the calculation node assigned the job It instructs the deletion of the common data and then prior to said job execution is completed in the calculation node assigned the job If the common data of the job that will be executed by the computing node that assigned the job matches, and is running the common data of the job execution is completed in the calculation node assigned the job compute nodes assigned the job If not at least for one of the case of using are for execution of other jobs, computer and delete instruction step, the instructing deletion of the common data of the job for the computing nodes assigned the job execution This is a job management method.

Job management apparatus described herein, the content of the job, execution of the said job information of the common data and said common data different from the individual data information corresponding to a plurality of jobs to be utilized in the job, it is included an acquisition unit for acquiring instruction, based on the execution command of the job acquired by the acquiring unit, allocating the job to one of the plurality of computing nodes, a plurality of jobs to compute node assigned the job transfer the common data and individual data used in the execution, and the execution instruction unit for instructing the execution of the job, with respect to computing node assigned the job execution is completed in the calculation node assigned the job the instructs the deletion of the job of the individual data, the common data of the job execution is completed in the calculation node assigned the job If then the common data of the job that will be run on the compute nodes assigned the job matches, and the common data of the job execution is completed in the calculation node assigned the job compute nodes assigned the job If not at least for one of the case of using the execution of other jobs that are running, and a deletion instruction unit for instructing the deletion of the common data of the job for the computing nodes assigned the job, the I have.

  The job management program, the job management method, and the job management apparatus described in the present embodiment have an effect that efficient job processing can be performed.

It is a figure showing roughly the composition of the information processing system concerning one embodiment. FIG. 2A is a hardware configuration diagram of the login node, and FIG. 2B is a hardware configuration diagram of the job management node. It is a functional block diagram of a job management node. FIG. 4A is a diagram illustrating an example of a job script used in an embodiment, and FIG. 4B is a diagram illustrating an example of a conventional job script. FIG. 5A is a diagram illustrating a scheduling table, and FIG. 5B is a diagram illustrating an example of a job processing state in each computation node based on the scheduling table in FIG. 5A. It is a flowchart (the 1) which shows a process of a job management node. 10 is a flowchart (part 2) illustrating processing of a job management node. FIG. 8A to FIG. 8C are flowcharts (part 1 to part 3) showing processing of the computation node. FIGS. 9A and 9B are flowcharts (No. 4 and No. 5) showing the processing of the calculation node. It is the table | surface which put together the necessity of transfer of the common stage in file in one Embodiment, and the necessity of deletion.

  Hereinafter, an information processing system according to an embodiment will be described in detail with reference to FIGS.

  FIG. 1 schematically shows a configuration of an information processing system 100 according to an embodiment. The information processing system 100 is a large-scale supercomputer system as an example, and includes a plurality of user terminals 10, a large-scale shared file system 80, and an information processing apparatus 20. The user terminal 10, the large-scale shared file system 80, and the information processing apparatus 20 are connected to the network 70.

  The user terminal 10 is a terminal on which a user performs various operations. The user terminal 10 logs into the information processing apparatus 20 (login node 22) or transmits a job execution command (job script) described by the user to the information processing apparatus 20. Or

  The large-scale shared file system 80 is a file system used by the information processing apparatus 20 (a login node 22 and a job management node 24 described later).

The information processing apparatus 20 includes a login node 22, a job management node 24 as a job management apparatus, calculation nodes 26 _{1 to} 26 _N (N is a natural number of 2 or more), and local disks 28 _{1 to} 28 _N. . Each of the nodes 22, 24, 26 _{1 to} 26 _N is connected to the internal network 30.

  The login node 22 is a node for executing user login processing, program development (source editing / compilation, etc.), and job operation (submission / deletion, etc.). As shown in FIG. 2A, the login node 22 includes a CPU 90, a ROM 92, a RAM 94, a network interface 96, and the like, and each unit is connected to a bus 98.

  The job management node 24 is a node that executes job management (acceptance, etc.), job scheduling (determination of a calculation node to which a job is allocated), and job monitoring (monitoring of the status of the calculation node and job). As shown in FIG. 2B, the job management node 24 includes a CPU 190, a ROM 192, a RAM 194, a network interface 196, a portable storage medium drive 199, and the like, and each unit is connected to a bus 198. In the job management node 24, the CPU 190 executes a program (job management program) stored in the ROM 192 or the like, or a program (job management program) read from the portable storage medium 191 by the portable storage medium drive 199. Thus, the function of each unit in FIG. 3 is realized.

  FIG. 3 shows a functional block diagram of the job management node 24. As illustrated in FIG. 3, the job management node 24 executes an acquisition program 40, a scheduling unit 42, a stage-in processing unit 44, a job processing instruction unit 45, a stage-out processing unit 46, and a deletion when the CPU 190 executes a program. The function as the instruction unit 48 is realized.

  The acquisition unit 40 acquires a job (job script) input from the user terminal 10 to the login node 22 from the login node 22. FIG. 4A shows an example of a job script. As shown in FIG. 4A, in the job script, an individual stage-in file (a file used by each job alone) is specified in “STDIN”. In addition, a common stage-in file (a file shared with other jobs) is specified in “STDIN_COMMON”. In FIG. 4A, “6” is designated as the required number of CPUs. FIG. 4B shows a conventional job script, but the common stage-in file (STDIN_COMMON) is not specified in the conventional job script.

Returning to FIG. 3, the scheduling unit 42 determines the calculation node 26 that processes the job acquired by the acquisition unit 40 based on the processing capability (number of CPUs) of the calculation nodes 26 _{1 to} 26 _N and the viewpoint of load distribution. Process (scheduling). Here, an example of the result of scheduling by the scheduling unit 42 is shown in the scheduling table of FIG. The scheduling table of FIG. 5A includes fields of “case number”, “job queue”, “common stage in file”, and “allocation destination calculation node”. A serial number is entered in the “case number” field. In the “job queue” field, a job identifier (ID) is input. In the “common stage in file” field, an identifier of the common stage in file used in each job is input. In the “assignment destination calculation node” field, information on a calculation node to which each job is assigned as a result of scheduling is input. As a result of this scheduling, in this embodiment, as shown in FIG. 5B, Job1 and Job2 are processed in parallel in the calculation node (compute-1), and Job3 and Job3 are calculated in the calculation node (compute-2). It is assumed that Job5 is processed continuously and Job4 is processed alone in the compute node (compute-5).

  The stage-in processing unit 44 executes stage-in processing based on the scheduling table in FIG. 5A and the job script in FIG. That is, the stage-in processing unit 44 applies to the local disk of the computation node to which the job and the individual stage-in file used in the job and the common stage-in file (stored in the large-scale shared file system 80) are assigned. And transfer (called file staging). The stage-in processing unit 44 does not transfer the common stage-in file when the same common stage-in file as the common stage-in file used in the job is stored in the local disk of the computation node. To do.

  The job processing instruction unit 45 issues a job processing instruction to the computation node that has performed the stage-in processing.

  The stage-out processing unit 46 instructs the computation node that has executed the job process to execute the stage-out process. Here, the stage-out process means a process of transferring a job process result (output file) from the local disk to the user terminal 10.

  The deletion instruction unit 48 deletes the individual stage-in file of the job for which the stage-out processing has been completed from the local disk. The deletion instruction unit 48 also refers to the scheduling table (FIG. 5A) to determine whether the common stage-in file of the job for which stage-out processing has been completed should be deleted from the local disk, and based on the determination result Execute the process. The details of whether or not to delete the common stage-in file will be described later.

Returning to FIG. 1, the calculation nodes 26 _{1 to} 26 _N are nodes that execute jobs (job scripts) under the instruction of the job management node 24. The computation nodes 26 _{1 to} 26 _N have a plurality of CPUs, and each CPU executes processing in accordance with an instruction from the job management node 24.

The local disks 28 _{1 to} 28 _N function as a file store area for storing data used by the computation nodes 26 _{1 to} 26 _N for job execution. Note that data (individual stage-in file, common stage-in file) is transferred from the large-scale shared file system 80 to the local disks 28 _{1 to} 28 _N by the job management node 24. In the following, when any one of the calculation nodes 26 _{1 to} 26 _N and the local disks 28 _{1 to} 28 _N is indicated, they are expressed as “calculation node 26” and “local disk 28”.

  Next, processing of the job management node 24 and the calculation node 26 in the present embodiment will be described in detail with reference to the flowcharts of FIGS.

  In the process of FIG. 6, first, in step S10, the acquisition unit 40 stands by until a job (job script) is received. Then, when the acquisition unit 40 receives (acquires) a job (job script), the process proceeds to step S12, and the scheduling unit 42 executes a scheduling process. By this scheduling process, the computation node 26 to which the job is assigned is determined. Here, a case where the user who has input the job script is “A” and the job ID of the input job is “12345” will be described.

  Next, in step S14, the stage-in processing unit 44 starts job execution. Next, in step S16, the stage-in processing unit 44 starts common stage-in transfer processing. In the common stage-in process, first, in step S18, the stage-in processing unit 44 determines whether there is a common stage-in file designation (STDIN_COMMON) in the job script. If the determination is negative, the process proceeds to step S32 (the common stage-in process is not performed). If the determination is affirmative, the process proceeds to step S20.

  When the process proceeds to step S20, the stage-in processing unit 44 determines whether the common stage-in file specified in the job script is being transferred to the same calculation node in another job. When judgment here is affirmed, it transfers to step S22. When the process proceeds to step S22, the stage-in processing unit 44 determines whether there is a common stage-in file transfer process completion notification (see step S30) regarding the other job. If the determination is negative, the process returns to step S20. If the determination is positive, the process proceeds to step S32 (the common stage-in process is not performed).

On the other hand, if the determination in step S20 is negative, the process proceeds to step S24, and the stage-in processing unit 44 transfers the common stage-in file to any of the computation nodes 26 _{1 to} 26 _N. The entity of the common stage-in file is, for example, arranged under / scratch / common / $ User / $ JOBID (in this embodiment, for example, /scratch/common/A/12345/commondata.dat). To do.

  Next, in step S26, the stage-in processing unit 44 creates a symbolic link to the common stage-in file. In this case, as described above, since the entity of the common stage in file is placed under / scratch / common / $ User / $ JOBID, a user-specified path (/ scratch) can be used so that the job can refer to the common stage in file. Create a symbolic link to (/) (eg /scratch/12345/commondata.dat). In this way, the stage-in processing unit 44 ends the common stage-in file transfer process (step S28).

  Next, in step S30, the stage-in processing unit 44 issues a common stage-in transfer process completion notification. Thereafter, the process proceeds to step S32.

  When the process proceeds to step S32 (when step S18 is denied, when step S22 is affirmed, or after step S30), the stage-in processing unit 44 executes the individual stage-in process. In this case, the stage-in processing unit 44 transfers the individual stage-in file to the local disk of the computation node 26 to which the job is assigned (for example, /scratch/12345/input.dat).

  Next, when the process proceeds to step S34 in FIG. 7, the job processing instruction unit 45 outputs a job processing instruction to the assignment destination calculation node defined in the scheduling table (FIG. 5A).

  Next, in step S <b> 36, the stage-out processing unit 46 outputs an instruction for stage-out processing to the calculation node 26.

  Next, in step S <b> 38, the deletion instruction unit 48 outputs an individual stage in file deletion instruction to the calculation node 26. Next, in step S40, the deletion instruction unit 48 starts the common stage-in file deletion process. In this case, the deletion instruction unit 48 determines whether or not jobs using the same common stage-in file are being executed in the same calculation node 26 in step S42. If the determination here is affirmed, all the processes in FIGS. 6 and 7 are terminated (that is, the common stage-in file is not deleted from the local disk of the computation node 26). The case where the determination in step S42 is affirmative is a case like compute-1 in FIG. 5B, where Job1 and Job2 use the same common stage-in file (Data1), and This is the case when Job2 is running when Job1 is completed.

  On the other hand, if the determination in step S42 is negative, the process proceeds to step S44, and scheduling processing is executed. In this case, the deletion instruction unit 48 specifies the assignment destination calculation node of the subsequent job.

  Next, in step S46, the deletion instruction unit 48 determines whether the subsequent job is assigned to the same calculation node. If the determination here is negative, that is, if a subsequent job is not assigned to the same computation node (when the common stage-in file is not used immediately after), the process proceeds to step S50. In step S50, the deletion instruction unit 48 outputs an instruction to delete the common stage-in file to the calculation node 26, and ends all the processes in FIGS.

  On the other hand, if the determination in step S46 is affirmative, that is, if the subsequent job is assigned to the same calculation node, the process proceeds to step S48. In step S48, it is determined whether or not subsequent jobs use the same common stage-in file. If the determination here is affirmed, all the processes in FIGS. 6 and 7 are terminated (that is, the common stage-in file is not deleted from the local disk of the computation node). The case where the determination in step S48 is affirmative is a case like compute-2 in FIG. 5B, and the same common stage-in file (Data2) is used immediately after the processing of Job3 is completed. This is a case where JOb5 processing is executed.

  On the other hand, if the determination in step S48 is affirmative, the process proceeds to step S50, and the deletion instruction unit 48 outputs a deletion instruction for the common stage-in file to the calculation node 26, and all of FIG. 6 and FIG. The process ends.

  As described above, in the processing of FIGS. 6 and 7, when job processing is completed, when another job using the same common stage-in file is executed on the same calculation node, and job processing is performed. When another job using the same common stage-in file is executed on the same computation node immediately after completion, the deletion instruction unit 48 does not delete the common stage-in file from the local disk (on the local disk). To leave). In this way, it is not necessary to transfer a common stage-in file for each job, and the efficiency of stage-in processing can be improved. FIG. 10 shows a summary of whether or not common stage-in files need to be transferred and whether or not they need to be deleted (added fields to the scheduling table of FIG. 5A). In this embodiment, as shown in FIG. 10, it can be said that the processing is made more efficient as much as the common stage-in file transfer field and the common stage-in file deletion field are “unnecessary”.

  Next, the processing of the calculation node 26 will be described with reference to FIGS. 8 (a) to 9 (b). 8A to 9B are processes performed in parallel. Each process is repeatedly executed as needed even if it is once completed.

(Processing in FIG. 8A)
In the process of FIG. 8A, the computation node 26 stands by until a common stage-in file transfer process is accepted in step S100 (until step S24 is performed in the flowchart of FIG. 6). In this case, the computation node 26 proceeds to step S102 when the transfer process is accepted, and executes the transfer process of the common stage-in file. Next, in step S104, the computation node 26 transmits a common stage in-file transfer process completion notification to the job management node 24 and the user terminal 10, and ends all the processes in FIG.

(Processing in FIG. 8B)
In the process of FIG. 8B, the calculation node 26 stands by until an instruction for job processing is received in step S110 (until step S34 is performed in the flowchart of FIG. 7). In this case, the computation node 26 proceeds to step S112 when executing the job processing instruction and executes the job processing. Next, in step S114, the computing node 26 transmits a job processing completion notification to the job management node 24 and the user terminal 10, and ends all the processing in FIG.

(Processing in FIG. 8C)
In the process of FIG. 8C, the computation node 26 stands by until an instruction for the stage-out process is received in step S120 (until step S36 is performed in the flowchart of FIG. 7). In this case, when the calculation node 26 receives an instruction for the stage-out process, the calculation node 26 proceeds to step S122 and executes the stage-out process. Next, in step S124, the computation node 26 transmits a stage-out process completion notification to the job management node 24 and the user terminal 10, and ends all the processes in FIG.

(Processing in FIG. 9A)
In the process of FIG. 9A, the computation node 26 stands by until an individual stage in-file deletion process instruction is accepted in step S130 (until step S38 is performed in the flowchart of FIG. 7). In this case, when the calculation node 26 receives an instruction for the individual stage in file deletion process, the calculation node 26 proceeds to step S132 and executes the individual stage in file deletion process. Next, in step S134, the computation node 26 transmits an individual stage in file deletion process completion notification to the job management node 24 and the user terminal 10, and ends all the processes in FIG.

(Processing in FIG. 9B)
In the process of FIG. 9B, the computation node 26 stands by until an instruction for the common stage-in file deletion process is received in step S140 (until step S50 is performed in the flowchart of FIG. 7). In this case, when the calculation node 26 receives an instruction for the common stage in file deletion process, the calculation node 26 proceeds to step S142 and executes the common stage in file deletion process. Next, in step S144, the computation node 26 transmits a common stage in-file deletion process completion notification to the job management node 24 and the user terminal 10, and ends all the processes in FIG. 9B.

  As is apparent from the above description, the function as the execution instruction unit is realized including the scheduling unit 42, the stage-in processing unit 44, and the job processing instruction unit 45.

  As described above in detail, according to the present embodiment, based on the job execution instruction acquired by the acquisition unit 40, the scheduling unit 42 assigns the job to one of a plurality of calculation nodes, and the stage-in processing unit 44. The common stage-in file and the individual stage-in file used for execution of each job are transferred to the computation node assigned by, and the job processing instruction unit 45 issues a job execution instruction. Then, when the common stage-in file of the job that has been executed and the common stage-in file of the job that is executed next on the same calculation node match (S46: affirmative, S48: affirmative) If the common stage-in file of the completed job is used for execution of another job being executed on the same calculation node (S42: Yes), the calculation node 26 To delete the common stage-in file. In this way, compared to the case where stage-in processing is performed for each job as in the past, and all of the stage-in files are deleted from the local disk of the computation node every time the job is completed, the common stage Since the number of in processes can be reduced, the process can be made more efficient. In addition, according to the present embodiment, it is possible to suppress the occurrence of performance degradation of the system itself due to delay in job turnaround time, cache overflow, and I / O contention.

  Further, according to the present embodiment, the deletion instruction unit 48 checks the scheduling table (FIG. 5A) when the job execution is completed, and identifies the assignment destination calculation node of the subsequent job. It is possible to appropriately issue an in-file deletion instruction.

  Further, according to the present embodiment, the stage-in processing unit 44 determines that the common stage-in file of the job is the common stage-in file of the job that was executed immediately before in the calculation node assigned to the job or the job that is currently being executed. If the common stage-in file is the same, the common stage-in file is not newly transferred to the computation node. As a result, it is possible to prevent the same common stage-in process from being performed repeatedly, so that the efficiency of the process can also be improved.

  In the above embodiment, the case where job scheduling is performed after job reception (S12) and before common stage-in file deletion (S44) after completion of stage out has been described. However, the present invention is not limited to this. . For example, the job may be scheduled after the job is completed, after the job status is changed, or after the setting of the scheduling unit 42 is changed, or the job may be scheduled periodically.

  In the above embodiment, the deletion instruction unit 48 matches the common stage-in file of the job that has been executed and the common stage-in file of the job that is executed next on the same calculation node (S46: Yes, S48: Yes), and the case where the common stage-in file of a job that has been executed is used to execute another job executed on the same computation node (S42: Yes) In this case, the case of deleting the common stage-in file has been described, but the present invention is not limited to this. For example, the common stage-in file may be deleted only when it corresponds to one of the above two cases.

  In the above embodiment, the case where the stage-in file used by the calculation node is transferred to the local disk of the calculation node has been described. However, the present invention is not limited to this. For example, when there is a file system shared between the computation nodes, the stage-in file may be transferred to the file system.

  The above processing functions can be realized by a computer. In that case, a program describing the processing contents of the functions that the processing apparatus should have is provided. By executing the program on a computer, the above processing functions are realized on the computer. The program describing the processing contents can be recorded on a computer-readable recording medium (except for a carrier wave).

  When the program is distributed, for example, it is sold in the form of a portable recording medium such as a DVD (Digital Versatile Disc) or a CD-ROM (Compact Disc Read Only Memory) on which the program is recorded. It is also possible to store the program in a storage device of a server computer and transfer the program from the server computer to another computer via a network.

  The computer that executes the program stores, for example, the program recorded on the portable recording medium or the program transferred from the server computer in its own storage device. Then, the computer reads the program from its own storage device and executes processing according to the program. The computer can also read the program directly from the portable recording medium and execute processing according to the program. Further, each time the program is transferred from the server computer, the computer can sequentially execute processing according to the received program.

  The above-described embodiment is an example of a preferred embodiment of the present invention. However, the present invention is not limited to this, and various modifications can be made without departing from the scope of the present invention.

In addition, the following additional notes are disclosed regarding the above description.
(Supplementary Note 1) Obtain a job execution instruction in which job details, common data information corresponding to a plurality of jobs used in the job, and individual data information different from the common data are described,
Based on the acquired job execution instruction, the job is assigned to one of a plurality of calculation nodes, and common data and individual data used for execution of each job are transferred to the assigned calculation node. I give an execution instruction,
When the common data of the job that has been executed in the predetermined calculation node matches the common data of the job that is executed next in the predetermined calculation node, and the common data of the job that has been executed in the predetermined calculation node If it is not at least one of the cases where it is used for execution of another job executed in the predetermined calculation node, an instruction to delete the common data is issued to the predetermined calculation node, and a process is executed on the computer Job management program
(Supplementary note 2) The supplementary note 1 is characterized in that, in the deletion process, each job assigned to each computation node is confirmed when execution of the job being executed in the predetermined computation node is completed. Job management program.
(Supplementary Note 3) In the process of issuing the instruction, when a job execution instruction is issued to a predetermined calculation node, the common data of the job is the common data of the job executed immediately before at the predetermined calculation node. The job management according to appendix 1 or 2, wherein the common data of the job for which the execution instruction is issued is not transferred to the predetermined calculation node when the common data of the job currently being executed is the same. program.
(Additional remark 4) The acquisition process which acquires the job execution instruction which described the content of the job, the information of the common data corresponding to the some job used by the said job, and the information of the separate data different from the said common data, ,
Based on the job execution instruction acquired in the acquisition step, the job is assigned to one of a plurality of calculation nodes, and common data and individual data used for execution of each job are transferred to the assigned calculation node. An execution instruction process for issuing a job execution instruction,
When the common data of the job that has been executed in the predetermined calculation node matches the common data of the job that is executed next in the predetermined calculation node, and the common data of the job that has been executed in the predetermined calculation node A deletion instruction step of issuing an instruction to delete the common data to the predetermined calculation node, unless it is at least one of the cases where it is used for execution of another job executed by the predetermined calculation node; A job management method executed by a computer.
(Additional remark 5) The said deletion instruction | indication process WHEREIN: When execution of the job currently performed in the said predetermined calculation node is completed, each job allocated to each calculation node is confirmed, The additional description 4 characterized by the above-mentioned. Job management method.
(Supplementary Note 6) In the execution instruction step, when a job execution instruction is issued to a predetermined calculation node, the common data of the job is the common data of the job executed immediately before in the predetermined calculation node or 6. The job management method according to appendix 4 or 5, wherein the common data of the job that issues the execution instruction is not transferred to the predetermined calculation node if the common data of the job that is currently being executed is the same .
(Additional remark 7) The acquisition part which acquires the execution instruction of the job in which the content of the job, the information of the common data corresponding to the some job used by the said job, and the information of the individual data different from the said common data are described ,
Based on the job execution instruction acquired by the acquisition unit, the job is assigned to one of a plurality of calculation nodes, and common data and individual data used for execution of each job are transferred to the assigned calculation node. An execution instruction unit for issuing a job execution instruction;
When the common data of the job that has been executed in the predetermined calculation node matches the common data of the job that is executed next in the predetermined calculation node, and the common data of the job that has been executed in the predetermined calculation node A deletion instructing unit that issues an instruction to delete the common data to the predetermined calculation node, unless it is at least one of the cases where it is used for execution of another job executed by the predetermined calculation node; A job management apparatus provided.
(Appendix 8)
The job management apparatus according to appendix 7, wherein the deletion instruction unit confirms each job assigned to each calculation node when execution of the job being executed in the predetermined calculation node is completed. .
(Appendix 9)
When the execution instruction unit issues a job execution instruction to a predetermined calculation node, the common data of the job is the common data of the job that was executed immediately before in the predetermined calculation node or is currently being executed. 9. The job management apparatus according to appendix 7 or 8, wherein the common data of the job for which the execution instruction is issued is not transferred to the predetermined calculation node if the common data of the job is the same.

40 acquisition unit 26 ₁ ~ 26 _N compute nodes 42 scheduling unit (a part of the execution instruction unit)
44 Stage-in processing unit (part of execution instruction unit)
45 Job processing instruction section (part of execution instruction section)
48 Deletion Instructing Section 24 Job Management Node (Job Management Device)

Claims (5)

It acquires the contents of the job, and information of the common data and said common data different from the individual data information corresponding to a plurality of jobs to be utilized in the job execution command of the job include,
Based on the acquired execution command of the job, assign the job to one of the plurality of computing nodes, common data and transfers the individual data used in the execution of a plurality of jobs to compute node assigned the job Te, and instructs execution of the job,
Against compute node assigned the job, instructs the deletion of the individual data of the job executed in the computing node assigned the job is completed, the job execution is completed in the calculation node assigned the job If the common data of the job that will be executed by the common data and then compute nodes assigned the job matches, and the common data of the job execution is completed in the calculation node assigned the job assigned the job If not at least for one of the case of using the execution of other jobs running on the compute nodes, and instructs deletion of the common data of the job for the computing nodes assigned the job, processes the A job management program to be executed by a computer. The process for instructing the deletion includes confirming each job assigned to each computation node when execution of the job being executed in the computation node to which the job is assigned is completed. The job management program described. In the process for instructing the execution of the job, when instructing execution of job to compute node assigned the job, common data of the job had been executed immediately before the calculation nodes assigned the job The common data of the job instructing the execution is not transferred to the calculation node to which the job is assigned when the common data of the job or the common data of the currently executed job is the same. Or the job management program according to 2. And contents of the job, and information of the common data and said common data different from the individual data information corresponding to a plurality of jobs to be utilized in the job, an acquisition step of acquiring an execution command of the job include,
Based on the acquired execution command of the job, assign the job to one of the plurality of computing nodes, common data and transfers the individual data used in the execution of a plurality of jobs to compute node assigned the job Te, an execution instruction step of instructing the execution of the job,
Against compute node assigned the job, instructs the deletion of the individual data of the job executed in the computing node assigned the job is completed, the job execution is completed in the calculation node assigned the job If the common data of the job that will be executed by the common data and then compute nodes assigned the job matches, and the common data of the job execution is completed in the calculation node assigned the job assigned the job If not at least for one of the case of using the execution of other jobs running on the compute nodes, deletion instruction step of instructing the deletion of the common data of the job for the computing nodes assigned the job And a job management method executed by the computer. And contents of the job, and information and the common data different from the individual data information of the common data corresponding to a plurality of jobs to be utilized in the job, an acquisition unit that acquires an execution command of the job include,
Based on the execution command of the job acquired by the acquiring unit, allocating the job to one of the plurality of computing nodes, common data and individual data used in the execution of a plurality of jobs to compute node assigned the job DOO transfer the, and execution instruction unit for instructing the execution of the job,
Against compute node assigned the job, instructs the deletion of the individual data of the job executed in the computing node assigned the job is completed, the job execution is completed in the calculation node assigned the job If the common data of the job that will be executed by the common data and then compute nodes assigned the job matches, and the common data of the job execution is completed in the calculation node assigned the job assigned the job If not at least for one of the case of using the execution of other jobs running on the compute nodes, deletion instruction unit that instructs the deletion of the common data of the job for the computing nodes assigned the job A job management apparatus.

Images