JP2001282551A - Job processor and job processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve a problem that performance is deteriorated when a job network has a large scale and it becomes complicated in a system where a plurality of jobs are automatically executed by the job network where the execution sequence and the execution time of the job are defined. SOLUTION: Job schedule information and various pieces of definition information area set to be common data. Thus, schedule processing themselves are executed in parallel by sharing them from the plural threads of a client processing part, a timer processing part, an agent processing part and a queue processing part.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a job processing apparatus and method for efficiently scheduling jobs in a system for processing a plurality of jobs defined in a job net format.

[0002]

2. Description of the Related Art A system for processing a job defined in a job network format is disclosed in JP-A-5-120.
37 and JP-A-8-44574.

In the job processing system described in Japanese Patent Application Laid-Open No. 5-12037, the load is concentrated on the "operation control controller" because a single "operation controller" existing in the system controls all jobs. It can be.

In the job processing method described in Japanese Patent Application Laid-Open No. Hei 8-44574, an "event management task" and an "execution management task" are provided for each user. The load concentrates on both tasks and can be a bottleneck. For example, in a computer equipped with four processors, when one user defines a job net including thousands of jobs, events such as the start and end of all jobs are sent to the “event management task”. In this method of performing centralized processing by one task (process), only one processor is used, multiplexing of schedule processing cannot be performed, and the other three processors are wasted.

[0005]

The above prior art does not consider the multiplexing of the schedule processing, and when processing a huge and complicated job net, it is not possible to utilize sufficient hardware resources, and the performance is deteriorated. There was a problem of inviting.

Further, when the size of a job net becomes large, it takes a very long time to search for a job to be executed next from the definition of the job net.

Further, when a job net is defined by a set of a plurality of child job nets, the definition information of all the child job nets is expanded when the main job net is executed. There is a problem that it takes a very long time to start job execution of the job net.

Further, in a system in which all processes are performed on one computer, there is a problem that the job processing itself is limited.

SUMMARY OF THE INVENTION It is an object of the present invention to multiplex the schedule processing itself of a job net, thereby maximizing the use of the hardware resources of the computer even in the processing of a huge and complicated job net, and avoiding performance degradation. It is in.

[0010]

In order to achieve the above object, according to the present invention, information such as job schedule information, queue information and job definition information is used as shared data, and a client processing unit (which processes input / output from a user). ), Execution of a timer processing unit (processing a job according to time and time), an agent processing unit (processing input / output with an agent that actually executes a job), a queue processing unit (processing a queue of a job waiting to be executed), etc. It can be shared by the processing unit that changes the definition and status of the inside job net. Further, each processing unit is realized by a thread that can be executed in parallel with each other. When each processing unit performs exclusive control for each information of the shared data, the schedule processing of the job net can be multiplexed. In particular, by activating a plurality of agent processing units and queue processing units, the multiplicity of the schedule processing is further increased.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention include the following. One is that job schedule information is held in a list structure, and entries of completed jobs are deleted in order, thereby shortening the search time for entries of a specific job.

Further, in a job net in which a plurality of child job nets (child job nets can be nested) are defined, the child job net satisfies the execution start condition (the execution of the preceding job is terminated). Unless the job definition information is read from the job definition information, the job schedule information creation is speeded up. When the child job net satisfies the execution start condition, necessary job information is read from the job definition information, and job schedule information is created. As a result, even when processing a huge job net, it is possible to prevent an abnormally long time from being executed until the first job execution.

Further, when the main job net is started, the delay monitoring definition information stored separately from the job definition information is checked, and the delay monitoring designation of the jobs, job nets and child job nets included in the main job net is performed. If so, monitoring information is extracted from a job for which monitoring is specified in advance at the time of job definition, and an entry is registered in the delay monitoring information, so that delay monitoring can be performed without expanding job definition information of a child job net or a grandchild job net.

Further, a client function for defining, referring to, and updating a job net, a server function for scheduling a job according to the definition of the job net and queuing an executable job, and a function for actually executing the job extracted from the queue are provided. Separated from the agent function to be executed,
It is now possible to exchange job definition information, execution results, and the like using network communication means such as a socket interface. Thus, each function can be executed on the same computer or on different computers, and the multiplicity can be set freely for each function.

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows a basic configuration of the present invention. here,
An example is shown in which two users separately define, execute, and reference a job, and execute a job using two queues (two queues are created in a queue information area) and three agents.

The user 1 (211) and the user 2 (212) define and refer to jobs and job nets, define calendars and other information, respectively, via the client process 1 (201) and the client process 2 (202). Reference, monitoring of the execution status of jobs and job nets, and the like. Client process 1 (201) and client process 2 (202) are server processes (1
00) and receive the result.

The server process (100) comprises a shared data section (150) and a plurality of threads sharing the shared data section (150). Client processing unit 1 (111), client processing unit 2 (112), timer processing unit (121), agent processing unit 1 (131), agent processing unit 2 (1
32), the agent processing unit 3 (133), the queue processing unit 1 (141), and the queue processing unit 2 (142) are threads in the server process (100), respectively, and execute the shared data unit (150) in the process. Access with exclusive control. Since the exclusion process is performed for each piece of information, for example,
The client processing unit 1 (111) executes the job definition information (1
55), the queue processing unit 1 (1
41) can access the queue information (153).

The client processing unit 1 (111) and the client processing unit 2 (112) receive requests from each client process (201, 202), and receive job definition information (155), calendar definition information (156),
Execution environment definition information (157), delay monitoring definition information (15
8) and the like. Furthermore, schedule information is expanded in the job schedule information area (151), a job is input to the queue information area (153), and a calendar information area (152) and a delay monitoring information area (154) are set.

The timer processing unit (121) is started periodically, and starts a job waiting for a specific time, monitors a delay, and the like.

The agent processing unit 1 (131), the agent processing unit 2 (132), and the agent processing unit 3 (1)
33) are created for the number of agents, each of which corresponds to one agent. The number of agent processing units is the same as the number of agents.
The reason why the number of client processing units is equal to the number of clients is to prevent the processing trigger from the agent or the client from being dropped.
Each agent processing unit is configured to execute each agent process (30
1, 302, and 303), information such as a normal end of a job, an abnormal end of a job, an error and a warning, and the like.
Submit a job.

The queue processing unit 1 (141) and the queue processing unit 2 (142) are activated when a job is entered into the queue in charge in the queue information area (153), and the assigned agent processes (301, 302,3
03) to execute the job. When jobs are submitted to the queue collectively, the queue processing unit (141,
142) performs the job processing to the end in the input order. Many-to-many connection between the queue processing unit and the agent process is possible. In the example of FIG. 1, the queue processing unit 1 (14
1) can request job execution from both the agent process 1 (301) and the agent process 2 (302). For example, by separately providing means for acquiring the remaining processing capacity of the computer in which each agent process exists, the load of job processing can be distributed.

Various definition information (155, 156, 157,
The entity of 158) exists in the job management database (400), and each thread accesses the database when referring to the information. However, for speeding up, a part of each definition information may be buffered as shared data. In addition to this, by storing the job execution result, log information, and the like in the job management database (400), it is possible to execute the job again, manage the generation, and the like.

Even when there is one agent, the agent processing unit, the queue processing unit, the client processing unit,
The timer processing unit forms a multi-thread.

Hereinafter, the embodiment of the present invention will be described in detail using a definition example of a job net. FIG. 2 shows a definition example of a job net. The job net A (500) includes eight jobs and one job net. Job A01 (501)
Is executed first when the job net A (500) becomes executable. For example, job net A (500)
Is defined to be executed at 11:00 am, at 11:00 am, the timer processing unit (1 in FIG. 1)
21) is a job schedule information area (15 in FIG. 1).
1), the job net A (500) is taken out, and the definition information of the job net A is stored in the job definition information area (1 in FIG.
55) and the job schedule information (15)
1), and submits the job A01 (501) to the queue. When the job A01 ends, the job A02 (50
2), job A03 (503), job A04 (50
4) becomes executable. When the jobs A02 and A03 are completed, the job A05 (505) becomes executable. When the job A04 ends, the job net B (60
0) becomes executable. The job net B includes four jobs. First, the job B01 (601) and the job B02 (602) can be executed. When the two jobs are completed, the job B03 (603) can be executed. When the job B03 ends, the job B04 (604)
Can be executed, and when the job B04 ends, the job net B (600) ends. Similarly, job A06
(506), job A07 (507), job A08
(508) is put into the queue and executed when all the preceding jobs are completed. It is the agent processing unit that submits the succeeding job to the queue upon completion of the job.

FIG. 3 shows a job schedule information area (15) immediately after the job net A (500) becomes executable.
The situation of 1) is shown. This is obtained by expanding the job schedule information by the client processing unit.

The job net start pointer (1000) and the job net end pointer (1001) point to the entry (1002) of the job net A. However,
It is assumed that there is no other job net being executed. The internal job pointer of the entry (1002) of the job net A points to the head of the entries of all the internal jobs expanded in the list format. The order of this job entry has nothing to do with the order of execution. Entries are created in the order read from the job definition information (155 in FIG. 1) and are connected to the list.
To explain this a little more, the job net entry and the job entry only indicate the following. That is, FIG. 3 merely shows that the job net A includes the jobs A01 to B indicated by the job entry, and the execution order of each job is indicated by the pointer of the next job. is not. For example, the area of the internal job in the entry of the job net A may be increased, and then the pointer may be set from the job A01 to the job net B. However, in this case, job net A
Has a variable length depending on the number of jobs included in the job net A, and handling is troublesome. In order to avoid this, the entries of each job are linked by a pointer from the job A01 to the job net B to indicate that the job is included in the job net A.

Each job entry (1101 to 110)
9) each has a preceding job pointer. For example, since job A01 (501 in FIG. 2) has no preceding job,
The preceding job pointer of the job entry (1101) of the job A01 is NULL. The preceding job of job A05 (505 in FIG. 2) is job A02 (50 in FIG. 2).
2) and job A03 (503 in FIG. 2),
The preceding job pointer of the job entry (1105) of the job A05 is the entry (1204) of the job A02 and the job A
03 indicates the head of the list of the entry (1205). Here, there is no horizontal order of the preceding job entry. The preceding jobs of job A05 are job A02 and job A03
It just indicates that At this point, job net B (600 in FIG. 2)
(500 in FIG. 2) is registered in the job entry as one job.

The job schedule information includes a list of time waiting jobs and time waiting job nets in addition to the above-described job development list.

When the development of the job net A is completed, an executable job is searched from the job net A. Examine the job entry in order from the beginning to the end, and if the preceding job pointer is NULL
The job of the entry L is determined to be executable, and is entered into the queue. If the execution time of the job net has been set, the timer processing unit expands the job net and submits the first executable job to the queue. If the job net is defined to be executed immediately according to the execution instruction from the client, the first executable job is put into the queue by the corresponding client processing unit. The queues are in the queue information area (153), and the head pointer (2001, 2002) and the tail pointer (201) of each queue are provided.
1, 2012) and a queued job entry (2101). In the state shown in FIG. 3, there is no preceding job only in the entry (1101) of the job A01. Here, it is assumed that queue 1 is defined as an attribute of job A01 (contained in “job A01 information”). Job A01
(2101) is connected to the end of the queue 1 as an executable job. When a job is input to the queue 1, the job is transmitted to the queue processing unit 1 (141 in FIG. 1), and the queue processing unit 1 sequentially transfers the jobs stored in the queue 1 to the agent. Here, the job A01
Is sent to the agent process 1 (301 in FIG. 1).

The processing up to the queue entry in FIG. 3 is performed by a timer processing unit (121 in FIG. 1) or a client processing unit (FIG. 1).
111 or 112). For example, when the execution start time is set in the job net A (500), the processing described in FIG. 3 is performed by the timer processing unit (121), and the immediate execution is performed by the client process (201 or 202 in FIG. 1). When requested, the same processing is performed by the client processing unit (111 or 112 in FIG. 1).

FIG. 4 shows the state of the job schedule information immediately after the completion of the job A01 (501 in FIG. 2). Referring to FIG. 4, the job entry of job A01 (1 in FIG. 3)
101) and the preceding job entry (1201, 1202, 1203 in FIG. 3) of job A01 has been deleted from each list (just remove the link without actually erasing the data). At this point, job A02 (110
2), job A03 (1103), job A04 (11
Since there is no preceding job in 04), these three jobs are put into the queue. Here, it is assumed that job A02 and job A03 are defined to be processed in queue 1 and job A04 is processed in queue 2. The jobs (2102, 2103, 2104) accumulated in the queue are taken out by the respective queue processing units (141 or 142 in FIG. 1), and the appropriate agent processes (301, 30 in FIG. 1).
2 or 303). As described above, when the information on the processed job is deleted, an executable job without a preceding job becomes visible.

The processing up to the queue input in FIG. 4 is performed by the agent processing unit 1 (131 in FIG. 1). Because the processing of job A01 is agent process 1
(301 in FIG. 1), and the processing result is transferred to the agent processing unit 1 (131 in FIG. 1).

FIG. 5 shows the state of the job schedule information immediately after the completion of the jobs A02 and A04.
The job entry of job A02 (1102 in FIG. 4) and the preceding job entry (1204 in FIG. 4) and the job entry of job A04 (1104 in FIG. 4) and the preceding job entry (1210 in FIG. 4) are deleted from each list. I have.
Here, since the job A03 has not been completed yet, it is not deleted from the list, and a flag indicating that the job A03 is being executed is set in the job A03 information of the job entry (1103) ("job A03 information" also has a flag area). The next job that can be executed at this point is job net B where all preceding jobs have been completed.
(1109). When the job net B becomes executable, the definition information of the job net is read out instead of being put into the queue, and is expanded as job schedule information.

In the processing of FIG. 5, the entry of the job A02 is deleted by the agent processing unit 1 (13 in FIG. 1).
1). However, it is assumed that the job A02 has been sent to the agent 1. The agent processing unit 3 (133 in FIG. 1) deletes the entry of the job A04 and expands the job net B.

It is assumed that the job A03 (503 in FIG. 2) is completed during the development of the job net B. FIG. 6 shows the state of the job schedule information area 151 after the completion of the job net B expansion processing and the job A03 end processing.

Referring to FIG. 6, the job entry of job A03 (1103 in FIG. 5) and the preceding job entry (1205 in FIG. 5) are deleted, and the entry (1003) of job net B and its internal job entry (from 1301) are deleted. 1304) and the preceding job entry (1401 to 1
403) has been added. Incidentally, a flag indicating that the job is being executed is set in the job net B information (1109) in FIG. At this time, the job A05 having no preceding job
(1105), job B01 (1301), and job B02 (1302) are put into the queue. As described above, the plurality of agent processing units execute the job schedule processing in parallel. That is, a plurality of threads are executed in parallel to improve processing performance. Specifically, deletion of an entry relating to a job that has been completed in the job schedule information area by sharing a plurality of agent processing units, expanding a child job net, and submitting a job to a queue are performed. A request is made to execute the job for the agent assigned by.

Similarly, FIG. 7 shows the state after completion of job A05, job B01, and job B02, ie, job A
This shows a state in which the entry relating to Job 05, the entry relating to Job B01, and the entry relating to Job B02 have been deleted. Here, there is no job preceding the job A06 and the job B03. FIG. 8 shows job A06 and job B03.
The state after the end, that is, the entry regarding the job A06,
This shows a state in which the entry related to job B03 has been deleted. Here, there is no job preceding job B4. FIG. 9 shows a state after the job B04 and the job net B are completed. Since there is no job preceding the job A07 and the job A08, and the job net B has ended, the pointer at the end of the job net points to the job net A again.

As described above, the information of the child job net in the job net is not expanded until the child job net becomes executable. However, the monitoring of the delay of the job in the child job net must be performed before the child job net is developed.

For example, the job A06 (506) shown in FIG. 2 has end delay information indicating that a warning is issued if execution is not completed by 11:55 am, and the job B03 (603) indicates "end by 11:15 am Warn if not activated "
It is assumed that start delay information is defined. If the respective delay information is set at the time of expanding the job net including the corresponding job, the delay monitoring of the job B03 is not set until the job net B (600 in FIG. 2) becomes executable. Job net B is 11:04 am
If it were possible to do so in 5 minutes, the warning would be delayed by more than 30 minutes. In order to prevent this problem, since the definition is performed for each job, only jobs that require delay monitoring when storing the job in the server are extracted, and the information is recorded as delay monitoring definition information (158 in FIG. 1). Then, when the main job net becomes executable, the definitions of the delay monitoring of all the jobs under the main job net are read from the delay monitoring definition information (158 in FIG. 1), and the delay monitoring information area (154 in FIG. 1). Set to. This is performed, for example, when the client processing unit develops the job net. The job execution monitoring is performed by the timer processing unit 121, for example, and a warning is issued when the job matches the delay monitoring information. Separately, a processing unit that monitors job execution may be provided. By doing so, delay monitoring can be performed without expanding all the jobs, so that an enormous amount of time is not required in the initial stage of execution.

FIG. 10 shows the state of the delay monitoring information area (154) immediately after the start of job net A. Entries of job A06 (3506) and job B03 (3603) for which delay monitoring is defined are registered. This is periodically referred to by the timer processing unit. In addition, temporarily
If delay monitoring information is added to the first job in the first parent job net, the amount of information alone increases and the handling becomes complicated in processing.

Referring to FIG. 1, the server process (10
0), client processes (201, 202), and agent processes (301, 302, 303) are independent processes. Communication between a client and a server, and communication between a server and an agent are performed by a general method such as a socket interface. This is performed using a simple communication means.
This provides flexibility in the configuration of the computer that executes each process.

FIG. 11 shows an example in which the client 1, the server, and the agent are all executed by the computer 1 (901). This configuration indicates that the entire process can be performed with only one computer. There may be multiple clients.

FIG. 12 shows agents (301, 302).
Only the other computers (902, 903) execute the configuration. In this configuration, since the agents are placed on different computers, the actual job processing is performed on another computer, and the load on the computer 1 (901) is reduced. Further, as shown in FIG. 13, it is also possible to execute all the processes on different computers. That is, only the server is computer 1
(901), Agent 1 (301), Agent 2 (302), and Agent 3 (303) are placed in Computer 2 (902), Computer 3 (903), and Computer 4 (904), respectively, and Client 1 (201). , The client 2 (202) with the computer 5 (905), respectively.
The computer is dedicated to the process in the computer 6 (906). By doing so, hardware resources can be used effectively.

[0044]

According to the present invention, since the schedule processing itself can be executed in parallel, even a job net in which a large number of jobs are defined in a complicated manner can be efficiently scheduled.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a basic configuration of an embodiment of the present invention.

FIG. 2 is a job net definition diagram according to the present invention.

FIG. 3 is a data configuration diagram of a job schedule information area and a queue information area immediately after execution of job net A in FIG. 2;

FIG. 4 is a data configuration diagram of a job schedule information area and a queue information area immediately after the end of job A01 in FIG. 2;

FIG. 5 is a data configuration diagram of a job schedule information area and a queue information area immediately after the end of jobs A02 and A04 in FIG. 2;

FIG. 6 is a data configuration diagram of a job schedule information area and a queue information area immediately after the completion of job A03 and the expansion of job net B in FIG. 2;

FIG. 7 shows job A05, job B01, and job B in FIG.
FIG. 11 is a data configuration diagram of a job schedule information area and a queue information area immediately after the end of 02.

FIG. 8 is a data configuration diagram of a job schedule information area and a queue information area immediately after the completion of a job A06 and a job B03 in FIG. 2;

9 is a data configuration diagram of a job schedule information area and a queue information area immediately after the end of job B04 and job net B in FIG. 2;

FIG. 10 is a data configuration diagram of a delay monitoring information area immediately after execution of job net A in FIG. 2;

FIG. 11 is a block diagram illustrating a computer configuration according to an embodiment of the present invention.

FIG. 12 is a block diagram illustrating a computer configuration according to another embodiment of the present invention.

FIG. 13 is a block diagram illustrating a computer configuration according to another embodiment of the present invention.

[Explanation of symbols]

 100. Server process 201. Client process 1 202 / client process 2 301. Agent process 1 302. Agent process 2 303. Agent process 3 150. Shared data section 111. Client processing unit 1 112. Client processing unit 2 121. Timer processing unit 131. Agent processing unit 1 132. Agent processing unit 2 133. Agent processing unit 3 141. Queue processing unit 1 142. Queue processing unit 2 400. Job management database 500. Job net A 600. Job net B

 ────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Masaaki Oya 5030 Totsuka-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture F-term in the Software Division of Hitachi, Ltd. (Reference) 5B098 GA03 GA05 GB01 GB11 GB13 GC01 GD02 GD14

Claims (6)

[Claims] 1. A job processing apparatus for sequentially executing a plurality of jobs by using a job net in which a job execution order is defined, comprising: a shared data unit having a job schedule information area for storing job schedule information; A job processing apparatus comprising: a plurality of threads that share a data portion, expand the job net in the job schedule information area, and refer to or update the expanded job schedule information. 2. The job schedule information developed in the job schedule information area includes a job entry storing job definition information of a job included in a job net, and a preceding job indicating a job to be executed prior to each job. At least one of said threads
2. The job processing apparatus according to claim 1, wherein an entry relating to the executed job is deleted. 3. The shared data section further has an area for storing a queue for job execution, and the thread refers to the expanded schedule information and should be executed next without a preceding job entry. A thread for selecting a job and submitting the job to the queue, and further comprising a thread which is activated when the job is submitted to the queue in charge and requests the assigned agent to execute the job. 2. Job processing device. 4. The job net according to claim 1, wherein said job net includes a child job net, and said child job net is expanded in said job schedule information area when said child job becomes executable. Job processing device. 5. A job processing apparatus for sequentially executing a plurality of jobs by using a job net in which a job execution order is defined, wherein job definition information defining a job and job processing delay included in the job definition information are defined. A storage device for storing delay monitoring definition information obtained by extracting information for specifying monitoring, and extracting the delay monitoring definition information of a job included in the main job net when the main job net is started, and a common data unit And a means for referring to the delay monitoring information to issue a warning. 6. A job processing method using a job net in which a job execution order is defined, wherein the job net includes an entry for storing definition information of each job included in the job net and the job entry. A step of expanding into a preceding job entry indicating a job to be executed prior to the job indicated by a pointer from the entry of each job of the entry, and a job having an entry having no preceding job entry as a job to be executed A job processing apparatus comprising: a step of selecting and submitting a job to a queue; and a step of deleting an entry relating to a job whose execution has been completed.