JP5445124B2 - Job flow management device, job flow execution start time prediction method and program thereof - Google Patents

Description

  The present invention relates to a job flow management device, a job flow execution start time prediction method, and a program thereof, and more particularly to a job flow management device, a job flow execution start time prediction method, and a job flow execution start time prediction method that can accurately predict the start time of each job in a job flow to be executed. Regarding the program.

  The job flow management apparatus is a job flow in which a plurality of batch jobs (hereinafter simply referred to as jobs) are defined in a flow form, and each computer is connected to another computer apparatus via a network in a predetermined order. It is a computer device that is transferred to and executed. Each job is described in a script language such as a UNIX (registered trademark) shell script or a Windows (registered trademark) batch script.

  For example, the processing based on this job flow is often used for applications such as batch execution of standardized batch processing on the computer at night or holidays when there are few users of the computer. In addition, there is a case where a target end time is set in the jobs constituting the job flow, and the job needs to be completed by the target end time. Specifically, the target end time is set when, for example, “It is necessary to complete the aggregation of data related to the previous day's sales by the start time of each day”.

  If such a job is not completed by the target end time due to an error or an excessive amount of data to be processed, the business of the organization will be delayed. Therefore, the execution start time and execution end time of a specific job are accurately predicted in the job flow, and if those times are likely to be delayed from the target time, the execution is accelerated by putting more computer resources into the job flow. It is important to take measures such as

  Regarding the prediction of the job flow execution start time or execution completion time, there are the following technical documents. Among these, the following Patent Document 1 describes a job progress monitoring system in which a monitoring schedule is updated using history information of executed jobs. Japanese Patent Application Laid-Open No. 2004-228561 describes a technique of processing a flow within a time limit even if a time setting related to a specific job is changed in a cooperative flow executed by a plurality of devices connected to a network. Patent Document 3 describes a process execution control system in which a process end time is predicted using the execution time of the same job in the past accumulated in execution log information.

  Patent Document 4 describes a job monitoring apparatus that obtains the latest end date and time of a job chain (job flow) and displays a list of delayed jobs. Patent Document 5 describes a job execution monitoring method that issues a warning when execution of a job flow is delayed beyond a preset allowable delay time. Patent Document 6 describes a technique of calculating the “margin” applied to the CPU when each task (job) is executed.

JP 2004-005205 A JP 2007-133696 A JP 2008-059470 A JP 05-346862 A Japanese Patent Laid-Open No. 08-095916 Japanese Patent No. 3660083

  In an actual job flow, there are various kinds of execution of each job constituting the job flow, such as merge processing (AND merge, OR merge, etc.), branch processing (AND branch, OR branch, etc.), waiting condition, exclusive constraint, etc. In many cases, restrictions are set. Patent Documents 1 to 6 do not disclose any processing in the case where these restrictions are set when executing each job. That is, when these restrictions are set, the execution start time cannot be accurately predicted by the techniques described in Patent Documents 1 to 6 or a technique that combines these techniques.

  The object of the present invention is to accurately predict an execution start time even when execution restriction conditions are set for each job constituting a job flow, thereby ensuring that a job whose execution is delayed It is an object of the present invention to provide a job flow management device, a job flow execution start time prediction method, and a program thereof that can be grasped.

  In order to achieve the above object, a job flow management apparatus according to the present invention transfers each job constituting a job flow to a preset execution apparatus, requests execution thereof, and sends the execution result from the execution apparatus. A job flow management apparatus for receiving, storage means for storing job attribute data indicating the attribute of each job in advance, and schedule management for instructing start of job flow execution according to a schedule preset in job attribute data Means, job flow definition interpretation means for calculating the expected start time of each job, and job submission for transferring each job to the execution apparatus, requesting execution, and receiving the execution result from the job execution apparatus Means and an event notification means for displaying the execution status of the job to the user. Alternatively, the job execution data is represented by each node indicating the confluence, and the job attribute data indicates the time constraint indicating the time when a specific job in the job is necessarily executed, and the target execution indicating the target time for starting the execution of each job. The job flow definition interpreting means includes the time and the expected execution time that is expected to be executed for each job, and the job flow definition interpreting means determines the execution order between each job that joins a specific junction in the job flow. Execution order determination unit that determines based on attribute data, and predicts the expected start time of the job sequentially from the job with the lower execution order using the time constraint and the expected execution time while tracing each node from the end point of the job flow When the calculated start time is delayed by a predetermined time or more compared to the target execution time, a warning is issued via the event notification means. And having a warning transmission unit for transmitting a.

  In order to achieve the above object, a job flow execution start time prediction method according to the present invention stores in advance job attribute data indicating attributes of each node constituting a job flow represented by each node indicating a job or a junction. In the job flow management device that transfers each job to a preset execution device and requests execution, and receives the execution result from the execution device, the job must be executed as job attribute data. Job attribute data is stored in advance, including data including a time constraint that indicates a power time, a target execution time that indicates a target time at which the job should be executed, and an expected execution time that is expected to be the job execution. According to the schedule set in advance, the schedule management means instructs the start of job flow execution, and the job attribute data Based on the job flow, the execution order determination unit determines the execution order in the order of execution that is the slowest among the jobs that merge at a specific merge point in the job flow. Using the time, the expected start time calculation unit calculates the expected start time of the jobs in order from the job whose execution order is late, and the calculated expected start time is compared with the target execution time preset in the job attribute data. The warning transmission unit issues a warning to the user when the time is more than the predetermined time, the job input means transfers each job to the execution device, requests execution, and receives the execution result from the execution device. It is characterized by that.

  In order to achieve the above object, the job flow execution start time prediction program according to the present invention indicates the attribute of each node constituting the job flow represented by each node indicating a job or a merge point, and always executes the job. Pre-store job attribute data including a time constraint that indicates a time to be executed, a target execution time that indicates a target of a time at which the job should be executed, and an expected execution time that is expected to be executed for the job, Each job is transferred to a preset execution device and requested to be executed, and the execution result is received from the execution device. According to a schedule preset in job attribute data Joining a specific merge point in the job flow based on the job attribute data, the procedure for instructing the job flow execution start A procedure for determining the execution order in the order of the latest execution among jobs, tracing each node from the end point of the job flow, and using the time constraint and the expected execution time, and the expected start time of the job sequentially from the job with the lowest execution order , A procedure for issuing a warning to a user when the calculated expected start time is delayed by a predetermined time or more compared to a target execution time preset in job attribute data, and an apparatus for executing each job And executing the procedure for receiving the execution result from the execution apparatus and causing the computer to execute the transfer.

  As described above, the present invention is configured to calculate the expected start time for each branch while tracing the node backward from the end point of the workflow. Therefore, even if there are restrictions such as branch processing, the execution order is appropriately determined. The expected start time can be calculated by this, so that even when execution constraints are set for each job, the execution start time can be accurately predicted, and the job that is delayed in execution can be assured. Therefore, it is possible to provide a job flow management apparatus, a job flow execution start time prediction method, and a program thereof having an excellent characteristic that it is possible to grasp the above.

It is explanatory drawing which shows the structure of the job flow execution system which concerns on the 1st Embodiment of this invention. It is explanatory drawing which shows an example of the job flow performed with the job flow execution system shown in FIG. FIG. 3 is an explanatory diagram in which the job flow shown in FIG. 2 is written in a list structure format. FIG. 3 is an explanatory diagram illustrating job attribute data indicating attributes of each node in the job flow illustrated in FIG. 2 and a list structure associated therewith. 5 is a flowchart showing processing performed by the job flow management apparatus shown in FIG. 1 for the job attribute data and list structure shown in FIG. It is explanatory drawing which shows the job attribute data after performing the process shown in FIG. It is explanatory drawing which shows the structure of the job flow execution system which concerns on the 2nd Embodiment of this invention. It is explanatory drawing which shows an example of the job flow performed with the job flow execution system shown in FIG. It is explanatory drawing which expressed the job flow shown in FIG. 8 in the list structure format. It is explanatory drawing which shows the job attribute data which shows the attribute of each node of the job flow shown in FIG. 8, and the list structure accompanying this. FIG. 8 is an explanatory diagram illustrating a program source list illustrating an example of processing performed by the job flow definition interpreting unit illustrated in FIG. 7 on a job flow. It is a continuation of FIG. It is a continuation of FIG. It is a continuation of FIG.

(First embodiment)
Hereinafter, the structure of the 1st Embodiment of this invention is demonstrated based on attached FIG.
First, the basic content of the present embodiment will be described, and then more specific content will be described.
The job flow management apparatus 11 according to this embodiment transfers each job constituting a job flow to a preset execution apparatus, requests execution thereof, and receives the execution result from the execution apparatus. Device. The job flow management apparatus 11 stores job attribute data 110 indicating attributes including time constraints and estimated execution times of each job in advance, and a job flow according to a schedule set in advance in the job attribute data. Schedule management means 101 for instructing the start of execution, job flow definition interpreting means 103 for calculating the expected start time of each job, each job is transferred to the job execution apparatus and requested for execution. A job input unit 104 that receives a result from the job execution device, and an event notification unit 105 that displays a job execution status to the user. Here, the job flow is represented by each node indicating a job or a confluence, and the job attribute data 110 includes a time constraint indicating a time when the job should be executed and a target execution time indicating a target of the time when the job should be executed. And an expected execution time that is an expected time for executing the job. The job flow definition interpreter 103 then determines an execution order between each job that joins a specific joining point in the job flow based on the job attribute data, and determines each node from the end point of the job flow. An expected start time calculation unit (CRST calculation unit 103a) that sequentially calculates an expected start time of a job in a slow execution order using time constraints and an expected execution time while tracing, and the calculated expected start time is a target execution And a warning transmission unit (CRST determination unit 103c) that issues a warning via the event notification means when the time is delayed by a predetermined time or more compared to the time.

  In addition, the expected start time calculation unit (CRST calculation unit 103a) is determined by the execution order determination unit when an exclusive condition that the job cannot be executed simultaneously with another specific job is set. Based on the execution order between the job and the job subject to the exclusion condition, the expected start time is calculated so that the execution time does not overlap with the job that is executed later in the job.

  Then, the expected start time calculation unit (CRST calculation unit 103a) is processed immediately after a job when a waiting condition is set for the job to start after another specific job has been executed. The expected start time is calculated based on the earlier time of the start time of the job and the start time of the job subject to the waiting condition.

With the above-described configuration, the job flow management apparatus 11 can accurately predict the execution start time even when execution constraint conditions are set.
Hereinafter, this will be described in more detail.

  FIG. 1 is an explanatory diagram showing a configuration of a job flow execution system 1 according to the first embodiment of the present invention. The job flow execution system 1 is configured by connecting respective computer devices of a job flow management device 11, a job execution device 12, and a job flow definition device 13 via a network 10.

  Although only one job execution device 12 and one job flow definition device 13 are shown in FIG. 1 for the sake of space, actually, a plurality of job execution devices are included in one batch job execution system 1. 12 may be included. Further, the job flow definition device 13 may be the same computer as the job flow management device 11 or the job execution device 12.

  The job flow management apparatus 11 includes a central processing unit (CPU) 21 that executes a computer program, a storage unit 22 that stores programs and data, and a communication unit that performs data communication with other computers via the network 10. 23. Similarly, the job execution apparatus 12 includes a CPU 31, a storage unit 32, and a communication unit 33.

  Similarly to these, the job flow definition device 13 includes a CPU 41, a storage unit 42, and a communication unit 43. In addition to this, the job flow definition device 13 displays a job flow processing result to the user and accepts an operation from the user. Output means 44 is provided.

  In the CPU 21 of the job flow management apparatus 11, a schedule management unit 101, a job flow definition storage unit 102, a job flow definition interpretation unit 103, a job input unit 104, and an event notification unit 105, which will be described later, are executed as computer programs. The storage means 22 of the job flow management apparatus 11 stores job attribute data 110 (to be described later) and a list structure 111 associated therewith. The CPU 31 of the job execution device 12 executes job execution means 151 and the CPU 41 of the job flow definition device 13 executes flow definition instruction means 201 as computer programs.

  The flow definition instruction unit 201 of the job flow definition apparatus 13 displays a GUI (Graphic User Interface) on the input / output unit 44, displays the job flow processing result to the user, and performs an operation from the user via this GUI. The input of job flow definition, job script, job attribute setting and the like is received, and the input contents are transferred to the job flow management apparatus 11.

  The schedule management unit 101 of the job flow management apparatus 11 instructs the job flow definition interpretation unit 103 to execute the stored job flow according to the job flow execution schedule stored in the job attribute data 110.

  The job flow definition storage unit 102 causes the job flow definition interpretation unit 103 to perform time setting analysis processing on the job flow definition transferred from the flow definition instruction unit 201 and then converts the job flow definition into job attribute data. Save as 110.

  The job flow definition interpreting unit 103 receives the instruction from the schedule management unit 101 and interprets the contents of the job flow, and then passes the job flow to the job input unit 104 for execution. Here, “interpreting the contents of a job flow” means calculating a CRST (Calculated Restriction on Start Time) that is a start time derived by calculation of each job (node) constituting the job flow, and calculating the calculated CRST. Is a process of determining whether or not there is a job that is delayed for a predetermined time or more.

  More specifically, the job flow definition interpreting unit 103 calculates a CRST calculation unit 103a that calculates a CRST, and an execution order determination unit 103b that determines the execution order of each node that joins a confluence on the job flow. The CRST determination unit 103c causes the event notification unit 105 to issue a warning when the CRST is delayed for a predetermined time or more.

  In response to an instruction from the job flow definition interpretation unit 103, the job input unit 104 of the job flow management device 11 transfers the contents of each job constituting the job flow to the job execution unit 151 of the job execution device 12, where Receive the results of each executed job. Then, the job flow execution result is notified to the job flow definition storage means 102 and stored as job attribute data 110.

  The event notification unit 105 of the job flow management apparatus 11 transfers the content of the event notified from the job flow definition interpretation unit 103 to the flow definition instruction unit 201 and causes the input / output unit 44 to display it.

  The job execution unit 151 of the job execution apparatus 12 executes each job transferred from the job input unit 104 of the job flow management apparatus 11 and returns the result to the job input unit 104.

  FIG. 2 is an explanatory diagram showing an example of a job flow 300 executed by the job flow execution system 1 shown in FIG. The job flow 300 includes 10 nodes, a node that performs jobs (batch processing) from “N1” to “N8”, two nodes that process flow merging, and an OR merging point “MP1”. , And an AND junction “MP2”.

  The OR merge point is a control for starting the processing of the subsequent node when at least one of the nodes indicated by the arrows toward the merge node is completed. If at least one of the “N3”, “N4”, and “N5” jobs is completed at the OR junction “MP1” of the job flow 300, the subsequent “N6” process can be started.

  The AND merge point is a control for starting the process of the subsequent node when all the processes of the node indicated by the arrow toward the node of the merge point are completed. The AND merging point “MP2” of the job flow 300 can start the subsequent “N8” process after both the “N6” and “N7” jobs are completed.

  In the job flow 300, a waiting condition is set in which “N7” waits for the completion of “N4”. Unless “N4” is completed, the process of “N7” is not started. Furthermore, exclusion restrictions are set for “N5” and “N7”, and it is impossible to execute the processes of “N5” and “N7” at the same time. When “N5” is started, “N7” starts after waiting for the completion of “N5”. Conversely, when “N7” is started, “N5” waits for the completion of “N7”. It will start from.

  Further, in the job flow 300, the restriction of the end time “X” is set for “N3”. A restriction on the start time “Y” is set for “N8”. That is, “N3” must be processed by time “X”, and “N8” must have started processing by time “Y”.

  FIG. 3 is an explanatory diagram in which the job flow 300 shown in FIG. 2 is written in a list structure format. FIG. 4 is an explanatory diagram showing job attribute data 110 indicating attributes of each node in the job flow 300 shown in FIG. 2 and a list structure 111 associated therewith. ID 110a is an identifier of each node corresponding to the display of FIG. The type 110b represents the type of each node. “Node” indicates a normal job (batch processing), “OrMeetingPoint” indicates an OR merging point, and “AndMeetingPoint” indicates an AND merging point.

  The wait target 110c has a value of 1 when the node is a node to be waited for in the wait condition, and a null value (represented by “−” in FIG. 4) otherwise. The queuing source list 110d stores a pointer to the list structure 111 that stores information on the standby target node when the node is a node that is a target of waiting under the waiting condition. When there are two or more target nodes, information of a plurality of nodes is stored in the list structure 111.

  The exclusive constraint 110e has a value of 1 when an exclusive constraint is set for the node, and a null value otherwise. The exclusive constraint list 110f stores pointers “Pt2” and “Pt3” to the list structure 111 indicating which node group has an exclusive relationship when the exclusive constraint 110e = 1.

  The start time (RST: Restriction on Start Time) 110g stores the time when the start time restriction is set for the node. If the start time constraint is not set, the value is null. As a result of analyzing the influence of the restriction of the start time 110g, the CRST 110h stores the restriction of the start time at the corresponding node derived by calculation. The CRST-ID 110i stores a value corresponding to the ID 110a of the node from which node the value of the CRST 110h is derived.

  The end time (RFT: Restriction on Finish Time) 110j stores the time when the end time restriction is set for the node. If the end time constraint is not set, the value is null. The estimated execution time (EET: Estimated Execution Time) 110k stores the time that is expected to be taken for execution of the job (batch processing) executed by the node. The estimated execution time of each job can use the actual value of the previous execution time of the same job, or can be calculated by applying any known technique.

  The previous node 110l stores a value corresponding to the ID 110a of the node processed before the node. However, when the node is a confluence, a plurality of nodes correspond to the previous node, and the previous node 110l is set to a null value. The list information 110m stores pointers “Pt4” and “Pt5” to the list structure 111 indicating the previous node when a plurality of nodes correspond to the previous node and the previous node 110l has a null value.

  The target time (TT: Target Time) 110n is a time set as a target value for the CRST 110h of each node. When the CRST 110h is delayed by a predetermined time or more compared to the target time 110n, the CRST determination unit 103c issues a warning via the event notification unit 105.

  FIG. 5 is a flowchart showing processing performed by the job flow management apparatus 11 shown in FIG. 1 for the job attribute data 110 and the list structure 111 shown in FIG. The job flow definition interpretation unit 103 that has received the flow definition storage instruction from the flow definition instruction unit 201 first checks whether or not the execution order determination unit 103b has a time restriction in the job flow (step S201). If there is no time restriction, there is no need for analysis, so the job flow definition storage unit 102 stores the job flow (step S208) and ends the process.

  If there is a time restriction in the job flow, the execution order determination unit 103b next checks whether there is an exclusive restriction in the job flow (step S202). If there is an exclusive constraint, the execution order determination unit 103b creates a list of nodes with the exclusive constraint (step S203), and the process proceeds to step S204. The exclusive node list is a list in which node information is stored in a list in the order in which jobs having exclusive execution relations are scheduled to be executed later. If there is no exclusion constraint, the process proceeds directly to step S204.

  Thereafter, the CRST calculation unit 103a calculates the start time constraint of each node (step S204). The start time constraints calculated in step S204 are CRST 110h and CRST-ID 110i shown in FIG. Then, for the node for which the end time 110j is set, the CRST determination unit 103c compares this with the calculated CRST 110h (step S205). When the CRST 110h is delayed by a predetermined threshold or more compared to the end time 110j, the CRST determination unit 103c notifies the event notification unit 105 of this and displays a warning to the user ( Step S206).

  Thereafter, the job flow definition storage unit 102 stores the job flow and the calculated start time constraint information as the job attribute data 110 and the list structure 111 (step S207), and ends the process.

  The calculation according to step S204 will be described in more detail using the examples shown in FIGS. In this calculation, processing is performed in the form of tracing each node from the end point of the job flow 300. First, “End → N8” is in the first line of FIG. 3, and “time X” is set as the start time 110g in the job attribute data 110 of FIG. Since the node “N8” is not affected by other time restrictions, “time X” is stored in the CRST 110h, and “N8” is stored in the CRST-ID 110i.

  In the second line of FIG. 3, “N8 → MP2” is indicated, and in the job attribute data 110, the previous node 110l of the node “N8” is “MP2”. Since the node “MP2” is an AND confluence, the previous node 110l has a null value, but the pointer “Pt5” to the list structure 111 indicating the previous node is stored in the list information 110m. When the pointer “Pt5” is referred to in the list structure 111, nodes “N6” and “N7” are stored.

  The previous nodes 110l of the nodes “N6” and “N7” are both “MP1”. Therefore, the job flow definition interpreter 103 calculates the CRST 110h of each node until the node “MP1” is reached. The CRST 110h of the node “N6” is a value obtained by subtracting EET6 that is the expected execution time 110k of the node “N6” from “time X” that is the CRST 110h of the node “N8”, that is, “X-EET6”. “N8” is stored in the CRST-ID 110i of the node “N6”.

  On the other hand, since the value of the exclusive constraint 110e is 1 for the node “N7”, it can be seen that the exclusive constraint is set. The CRST calculation process is performed from the node “N8”, which is the end point of the job flow 300, while tracing the previous node 110l of each node. The value of the exclusive constraint list 110f of the node “N7” is “Pt3”. When the pointer “Pt3” is referred to in the list structure 111, it can be seen that the node “N5” is the target of this exclusive process.

  In the job flow 300, the node “N5” is executed before the node “N7”, and the CRST 110h of the node “N5” is a null value, that is, has not yet been calculated. “N7” will be the slowest. Therefore, the CRST 110h of “N7” is calculated.

  The CRST 110h of the node “N7” is a value obtained by subtracting EET7 that is the expected execution time 110k of the node “N7” from “time X” that is the CRST 110h of the node “N8”, that is, “X-EET7”. “N8” is stored in the CRST-ID 110i of the node “N7”.

  Subsequently, the CRST-ID 110i of the node “MP1” that is the preceding node of the nodes “N6” and “N7” is analyzed. The node “MP1” is a branch point to the processing of the node “MP2” which is an AND junction. Therefore, it is necessary to proceed to the next node after all the jobs of the nodes “N6” and “N7” indicating the node “MP2” are completed. That is, it is necessary to use the earlier one of “X-EET6” that is the node “N6” CRST110h and X-EET7 that is the CRST110h of the node “N7” for monitoring.

  Here, when a minimum value among a plurality of times 1, time 2... Is expressed as MIN (time 1, time 2...), CRST 110h of the node “MP1” is MIN (X-EET6, X-EET7). It is. “N8” is stored in the CRST-ID 110i of the node “MP1”.

  Since the node “MP1” is an OR branch point, the pointer “Pt4” stored as the list information 110m of the node is referred to from the list structure 111. The pointer “Pt4” points to the nodes “N5”, “N3”, and “N4”. Then, these paths are matched for the first time at the node “N1” that is the starting point of the job flow 300. Accordingly, the CRST 110h of each node is calculated on each route from the node “MP1” to “N1”.

  Since the node “N5” has the exclusive constraint 110e = 1, it can be seen that the exclusive constraint is set. The exclusive restriction list 110f of the node “N5” indicates the pointer “Pt2”, and from this point, the node “N5” becomes an exclusive restriction with the node “N7”. The order of execution expected is “N5” and “N7”.

  Accordingly, the CRST 110h of the node “N5” is the earlier of the CRST of the node “MP1” and the CRST of the node “N7”, that is, MIN (MIN (X-EET6, X-EET7), X-EET7) − It is a value expressed in EET5. “N8” is stored in the CRST-ID 110i of the node “N5”.

  The CRST 110h of the node “N2” is a value obtained by subtracting EET2 from the CRST 110h of the node “N5”, that is, a value expressed by MIN (MIN (X-EET6, X-EET7), X-EET7) -EET5-EET2. .

  Further, the end time 110j is set for the node “N3”. Therefore, the CRST 110h of the node “N3” is calculated based on the end time 110j and the CRST 110h of the node “MP1” with the earlier start time condition.

  Therefore, the CRST 110h of the node “N3” is a value expressed by MIN (Y, MIN (X-EET6, X-EET7)) − EET3 when the end time 110j of the node is time Y. In the CRST-ID 110i of the node “N3”, “N3” or “N8” is stored based on the earlier start time condition adopted.

  Since the waiting target 110c is 1 for the node “N4”, it is understood that the node “N4” is a target node for the waiting condition. Since the pointer “Pt1” to the list structure 111 is stored in the waiting source list 110d of this node, when this is confirmed in the list structure 111, the standby target node is “N7”. The CRST 110h of the node “N4” is calculated based on the earliest start time of the nodes “N7” and “MP1”.

  Therefore, the CRST 110h of the node “N4” is calculated by MIN (X-EET7, MIN (X-EET6, X-EET7)) − EET4, and “N8” is stored in the CRST-ID 110i of the N4.

  The node “N1” is also a branch point to the processing of the node “MP1” that is the OR junction. Since the node “MP1” is an OR merging point, if any one of “N3”, “N4”, and “N5” jobs indicating this node is completed, the process proceeds to the subsequent node. Therefore, the latest time condition among the CRSTs “N2”, “N3”, and “N4” can be used for monitoring.

  Here, if the maximum value among time 1, time 2... That is a plurality of times is expressed as MAX (time 1, time 2...), CRST 110 h of node “N 1” is MAX (MIN (MIN (MIN (X-EET6, X-EET7), X-EET7) -EET5-EET2, MIN (Y, MIN (MIN (X-EET6, X-EET7))-EET3, MIN (X-EET7, MIN (X-EET6, X-EET7)) ) −EET4) −EET1 “N3” or “N8” is stored in the CRST-ID 110i of the node “N1” depending on the value adopted when taking the maximum value.

  FIG. 6 is an explanatory diagram showing the job attribute data 110 after the processing shown in FIG. 5 is performed. The CRSTs 110h of the nodes “N1” to “N8” are obtained according to the formulas described outside the table for the sake of space.

(Overall operation of the first embodiment)
Next, the overall operation of the above embodiment will be described. In the job flow execution start time prediction method according to the present embodiment, job attribute data indicating attributes of each node constituting a job flow represented by each node indicating a job or a junction is stored in advance. The job flow management apparatus 11 that transfers to a predetermined execution device and requests execution and receives the execution result from the execution device. Pre-stores data including constraints, a target execution time that indicates the target of the time at which the job should be executed, and an expected execution time that is expected to be the job execution, and is preset in the job attribute data The schedule management means instructs the start of job flow execution according to the schedule, and the job The execution order determination unit determines the execution order in the order of execution that is the slowest among the jobs that join a specific merge point in the flow, and uses the time constraints and expected execution time while tracing each node from the end point of the job flow. Then, the expected start time calculation unit sequentially calculates the expected start time of the jobs from the job whose execution order is late (FIG. 5: Step S204), and compares the calculated expected start time with the target execution time set in advance in the job attribute data. When the delay is over a predetermined time, the warning sending unit sends a warning to the user (FIG. 5: steps S205 to 206), and the job input means transfers each job to the execution device and requests execution. In addition, the execution result is received from the execution device.

Here, each of the above-described operation steps may be programmed to be executable by a computer, and may be executed by the job flow management apparatus 11 which is a computer that directly executes each of the steps.
With this configuration and operation, the present embodiment has the following effects.

  The present embodiment is configured to calculate the start time constraints sequentially from the node having the slowest execution order between the nodes having the exclusive constraints of the job flow in which many exclusive constraints are set. In addition, when the CRST value is not calculated in a node whose execution order is later, the CRST value is calculated so that the execution time does not overlap in the corresponding node after calculating the value. Therefore, it is possible to appropriately calculate CRST in accordance with the restriction of the start time at each node even for a flow having an exclusion restriction.

  Further, in this embodiment, when a waiting condition is set, the CRST value of the node waiting at the target node to be waited is compared with the CRST value of the subsequent node of the target node to be waited for calculation. Since it is configured, the CRST can be appropriately calculated according to the restriction of the start time at each node even for the flow for which the waiting condition is set.

(Second Embodiment)
In addition to the configuration of the job flow management apparatus 11 according to the first embodiment described above, the job flow management apparatus 411 according to the second embodiment of the present invention is replaced with a merging point of each node configuring the job flow. The expected start time calculation unit (CRST calculation unit 503a) of the job flow definition interpreting unit 503 traces each node from the start point to the end point of the job flow, and sets the node with the time constraint. When it arrives, it uses the expected execution time to calculate the expected start time of the job corresponding to the node, and the job corresponding to the other node while tracing the job flow in reverse from the calculated expected start time of the node And a function for calculating the expected execution time of the system.

With this configuration, the same effect as that of the first embodiment can be obtained for a job flow represented by a job and a branch point.
This will be described in more detail below.

  FIG. 7 is an explanatory diagram showing the configuration of the job flow execution system 401 according to the second embodiment of the present invention. The job flow execution system 401 is configured by connecting the respective computer devices of the job flow management device 411, the job execution device 12, and the job flow definition device 13 via the network 10. The job execution device 12 and the job flow definition device 13 are the same as those in the first embodiment.

  Similar to the job flow management apparatus 11 according to the first embodiment shown in FIG. 1, the job flow management apparatus 411 includes a CPU 21, a storage unit 22 that stores programs and data, and other computers via the network 10. And communication means 23 for performing data communication.

  In the CPU 21 of the job flow management apparatus 411, the schedule management unit 101, the job flow definition storage unit 102, the job flow definition interpretation unit 503, the job input unit 104, and the event notification unit 105 are executed as computer programs. The storage means 22 of the job flow management apparatus 11 stores job attribute data 510 (to be described later) and a list structure 511 associated therewith. The only difference from the job flow management apparatus 11 shown in FIG. 1 is the operation of the job flow definition interpreting means 503, the contents of the job attribute data 510 and the list structure 511, and everything else is the same.

  The job flow definition interpreting unit 503 includes a CRST calculation unit 503a that calculates CRST, an execution order determination unit 503b that determines the execution order of each node that joins a merge point on the job flow, and the calculated CRST is equal to or longer than a predetermined time. In the case of delay, the event notification unit 105 is divided into a CRST determination unit 503c that issues a warning. The sharing of these functions is in principle the same as the job flow definition interpreting means 103 in the first embodiment shown in FIG.

  FIG. 8 is an explanatory diagram showing an example of a job flow 600 executed by the job flow execution system 401 shown in FIG. The job flow 600 includes 10 nodes, a node that performs jobs (batch processing) from “N1” to “N8”, two nodes that process the merging of flows, and an OR branch point “BR1”. AND branch point “BR2”.

  In the job flow 600, the jobs “N1” to “N8” are the same as the job flow 300 executed in the first embodiment, and the waiting conditions, exclusive constraints, and end time constraints set for each job are the same. , And the contents of the start time constraints are also the same. However, the job flow 300 is composed of the jobs “N1” to “N8”, the OR merge point “MP1”, and the AND merge point “MP2”, whereas the job flow 600 is composed of the jobs “N1” to “N8”. ”, An OR branch point“ BR1 ”, and an AND branch point“ BR2 ”.

  The OR branch point performs control to start processing of the node at the merge point when at least one of the nodes indicated by the arrows toward the node at the merge point is completed when the branched processes merge into one node. The OR branch point “BR1” of the job flow 600 is a successor job of “N3”, “N4”, and “N2”, and merges when at least one of the “N5” jobs that merge with other jobs is completed. The subsequent “BR2” process can be started.

  An AND branch point is a control that starts processing of a node at a merge point when all of the processes of the node indicated by the arrow toward the node at the merge point are completed when the branched processes merge into one node. Do. The AND branch point “BR2” of the job flow 600 can start the processing of “N8” after joining after waiting for completion of the jobs “N6” and “N7”.

  FIG. 9 is an explanatory diagram in which the job flow 600 shown in FIG. 8 is written in a list structure format. FIG. 10 is an explanatory diagram showing job attribute data 510 indicating attributes of each node of the job flow 600 shown in FIG. 8 and a list structure 511 associated therewith. Since the contents of the job attribute data 510 and the list structure 511 are approximately the same as the job attribute data 110 and the list structure 111 shown in FIG. 4, the names and reference numbers are the same except for the differences described below.

  In the job attribute data 510, the type 510b represents the type of each node. “Node” indicates a normal job (batch processing), “OrBrunch” indicates an OR branch point, and “AndBrunch” indicates an AND branch point. The subsequent node 510l stores a value corresponding to the ID 110a of the node processed after the end of the node. However, if the node is a branch point, a plurality of nodes correspond to the subsequent nodes, and the subsequent node 510l is set to a null value. The list information 510m stores pointers “Pt4” and “Pt5” to the list structure 511 indicating the subsequent nodes when a plurality of nodes correspond to the subsequent nodes and the previous node has a null value.

  The job flow time setting analysis processing and the job flow storage processing performed by the job flow definition interpretation unit 503 for the job attribute data 510 and the list structure 511 of the job flow 600 are the same as those in the flowchart shown in FIG. It is.

  11 to 14 are explanatory diagrams showing a program source list showing an example of processing performed by the job flow definition interpreting unit 503 shown in FIG. 7 for the job flow 600 shown in FIG. This source list is written based on the C language, but omits the strict definition of the function type and “;” at the end of the line for the sake of simplicity. In the lists of FIGS. 11 to 14, the part enclosed by “//” is a comment line not directly related to the process, but a comment describing the actual process performed in that part is shown. doing. Any other programming language can be used to describe the program.

  In the source list of FIGS. 11 to 14, each node constituting the job flow is a “Node” structure, and among them, the node currently treated as a target of the arithmetic processing is “CurrentNode”, and the previous node is “PreviousNode”. Is defined. Further, the “Node” structure has the attributes shown in FIGS. 3 and 10. ID 110a is “Node.ID”, type 510b is “Node.Type”, meeting target 110c is “Node.WaitingTarget”, meeting source list 110d is “Node.WaitingSourceList”, exclusion constraint 110e is “Node.Exclusion”, exclusion constraint The list 110f is “Node.ExclusionList”, the start time 110g is “Node.RST”, the CRST 110h is “Node.CRST”, the CRST-ID 110i is “Node.CRST_ID”, the end time 110j is “Node.RFT”, and the expected execution time. 110k is represented as “Node.EET”, and the list information 510m is represented as “Node.SequenceList”. The same applies to “CurrentNode” and “PreviousNode”. When moving the processing target to the succeeding node, “Node-> Next” is set.

  In addition, in the source lists of FIGS. 11 to 14, list structures of “ExclusionList” and “WaitingSourceList” are defined. The former stores a node having an exclusive relationship, and the latter stores a node to be awaited, and the node and a list (pointer) following the node.

  The job flow definition interpretation unit 503 performs a process of recursively calculating the start time constraint from the first node of the job flow. First, execute the argument “N1” for the function GetCRST (Node).

  Since the node N1 designated as the argument “N1” has a subsequent job 510l, UseNextCRST (Node) is called, and a function GetCRST (BR1) for obtaining the start time constraint of the subsequent node 510l of the node “N1”, that is, the node “BR1” ), The start time constraint of the node “N1” is calculated by “GetCRST (BR1) −EET1” obtained by subtracting the estimated execution time EET1 of the node “N1”.

  Since the type 510b of the node “BR1” is “OrBrunch”, that is, an OR branch point, GetMaxCRST (SequenceList) is called. GetMaxCRST is a function that calls GetCRST (Node) for each node stored in the list structure 511 and uses the maximum value as a return value. In the list structure 511, the nodes “N2”, “N3”, and “N4” are stored in the pointer “Pt4” of the node “BR1”. Therefore, GetCRST (N2) and GetCRST (N3 ), GetCRST (N4) is called.

  In the node “N3”, a value obtained by subtracting EET3, which is the expected execution time 110 k of the node “N3”, from the start time constraint GetCRST (BR2) of the subsequent node 510l is written as CRST110h. Here, the end time is written in the node “N3” itself. An end time constraint Y is set as 110j. For this reason, processing for the end time constraint is performed.

  If the time obtained by subtracting EET3 from the end time constraint Y is smaller than GetCRST (BR2) −EET3, the end time constraint is a more severe condition. In this case, UseCurrentNodeRFT (N3) is called and the time is overwritten on the CRST 110h of the node “N3”.

  On the other hand, since node “N4” is a node to be waited for, GetMinCRST_Waiting (N4) is called. In this function, the queuing source list 110d of the node “N4” is extracted, and GetCRST (N7) is called for “N7” that is the queuing target of the node “N4”. In node “N7”, an exclusion constraint 110e with node “N5” is set, so GetCRST_Exclusion (N7) is called.

  The CurrentNode, which is a structure indicating the node being processed, corresponds to the node “N7” whose execution order is late, and since the CRST 110h of the node “N7” is a null value, the exclusive constraint 110e of the node “N7” is 0. And GetCRST (N5) is called again. In the process of GetCRST (N7), since the exclusive constraint 110e is set to 0 this time, GetCRST (N8) -EET7 is calculated.

  Since the start time 110g is set in the node “N8” and the CRST 110h is a null value, UseCurrentNodeRST (N8) is called and the time X is set in the CRST 110h. Thereby, the CRST 110h of the node “N7” is calculated as X-EET7, and the process returns to GetCRST_Exclusion (N7). In GetCRST_Exclusion (N7), both CurrentNode and Node are “N7” and match, so the processing with the same function is completed here, and the processing returns to GetMinCRST_Waiting (N4).

  In GetMinCRST_Waiting (N4), the CRST 110h of the node “N7” and the CRST 110h of the node “BR2” are compared, and a value obtained by subtracting EET4 from the smaller value is set as the CRST 110h of the node “N4”.

  Since node “N2” has subsequent node 510l “N5” up to node “BR1”, GetCRST (N5) −EET2 is calculated as CRST value as CRST110h. In the process of GetCRST (N5), since the node “N5” has the exclusion constraint 110e with the node “N7”, GetCRST_Exclusion (N5) is called. The node “N7” is initially set as the CurrentNode, but the CRST 110h of the node “N7” is already set, and the nodes pointed to by the structures of the CurrentNode and the Node are different, so the PreviousNode is set as the node “N7”. , The current node is processed to be the node “N5”.

  After this process, since the nodes indicated by the structures of CurrentNode and Node match, calculation is performed for the node that is the target of time constraint confirmation. Specifically, the exclusive constraint 110e of the node “N5” becomes null, and the CRST 110h is compared between the node “BR2” and the node “N7” that are the subsequent nodes 510l of the node “N5”, and the EET5 is calculated from the smaller value. The value obtained by subtracting CRST 110h of the node “N5”.

  The node “BR2” calculates GetCRST (BR2), but since BR2 is an And branch point, GetMinCRST (SequenceList) is called. GetMinCRST is a function that calls GetCRST (Node) for each node stored in the list structure 511 and uses the minimum value as a return value.

  In the list structure 511, since the nodes “N6” and “N7” are stored in the pointer “Pt5” of the node “BR2”, GetCRST (N6) and GetCRST (N7) are called for each of these nodes. The minimum value is selected from these values. Since GetCRST (N7) has already been calculated, GetCRST (N6) is calculated by “GetCRST (BR1) −EET6” obtained by subtracting the estimated execution time EET6 of the node “N6” from the CRST110h of the node “N8”. The CRST 110h of the node “BR2” is assumed.

  With the calculations so far, GetCRST (N8), GetCRST (N6), and GetCRST (N7) have been obtained with specific numerical values. Each of the above-described numerical values calculated as GetCRST (Node) is embodied by this, and up to GetCRST (N1) is calculated as a specific numerical value. Finally, the CRSTs 110h of the nodes “N1” to “N8” are obtained as the same values in the same manner as in the first embodiment shown in FIG.

  As described above, in the present embodiment, the job flow expressed by the branch point can be processed in the same manner as the process described in the first embodiment, and the same effect can be obtained. In other words, it is possible to calculate CRST in consideration of exclusive restrictions and waiting conditions.

  The present invention has been described with reference to the specific embodiments shown in the drawings. However, the present invention is not limited to the embodiments shown in the drawings, and any known hitherto provided that the effects of the present invention are achieved. Even if it is a structure, it is employable.

  The present invention can be applied to a job flow execution system.

DESCRIPTION OF SYMBOLS 1,401 Job flow execution system 10 Network 11, 411 Job flow management apparatus 12 Job execution apparatus 13 Job flow definition apparatus 21, 31, 41 CPU
22, 32, 42 Storage unit 23, 33, 43 Communication unit 44 Input / output unit 101 Schedule management unit 102 Job flow definition storage unit 103, 503 Job flow definition interpretation unit 103a, 503a CRST calculation unit 103b, 503b Execution order determination unit 103c , 503c CRST determination unit 104 Job input means 105 Event notification unit 110, 510 Job attribute data 110a ID
110b, 510b type 110c meeting target 110d meeting source list 110e exclusion constraint 110f exclusion constraint list 110g start time 110h CRST
110i CRST-ID
110j End time 110k Expected execution time 110l Previous node 110m, 510m List information 110n Target time 111 List structure 151 Job execution means 201 Flow definition instruction means 300, 600 Job flow 510l Subsequent node

Claims (6)

A job flow management device that transfers each job constituting a job flow to a preset execution device, requests execution thereof, and receives the execution result from the execution device,
Storage means for storing job attribute data indicating the attribute of each job in advance, schedule management means for instructing start of execution of the job flow according to a schedule preset in the job attribute data, and each of the jobs Job flow definition interpreting means for calculating the expected start time, job input means for transferring each job to the execution apparatus, requesting execution, and receiving the execution result from the job execution apparatus, And an event notification means for displaying the job execution status to the user,
The job flow is represented by each node indicating a job or a confluence,
The job attribute data includes a time constraint that indicates a time at which a specific job in the job is necessarily executed, a target execution time that indicates a target of a time at which the execution of each job is started, and the execution of each job And the expected execution time, which is the time expected to take
The job flow definition interpreting means
An execution order determination unit that determines an execution order between each job that joins the specific joining point in the job flow based on the job attribute data, and the time constraint while tracing each node from the end point of the job flow. And an expected start time calculation unit that sequentially calculates the expected start time of the job from the job with the slower execution order using the estimated execution time, and the calculated expected start time is compared with the target execution time. A job flow management apparatus, comprising: a warning transmission unit that transmits a warning through the event notification means when a predetermined time or more is delayed.   The exclusive start time calculation unit and the job set in the job attribute data when the exclusion condition that the job cannot be executed simultaneously with another specific job is set. Based on the execution order with the job subject to the condition, the expected start time is calculated so that the execution time does not overlap from the job that is executed later in the corresponding job, The job flow management apparatus according to claim 1.   The job start time to be processed immediately after the job is set when the expected start time calculation unit sets a waiting condition for the job to start after the execution of another specific job is completed. The job flow management apparatus according to claim 1, wherein the expected start time is calculated based on an earlier time among start times of jobs subject to the waiting condition. Instead of the merging point of each node constituting the job flow, it has a configuration equipped with a branch point,
The predicted start time calculation unit traces each node from the start point to the end point of the job flow, and when it reaches the node for which the time constraint is set, the job corresponding to the node using the predicted execution time And a function of calculating the expected execution time of the job corresponding to the other node while tracing the job flow in reverse from the calculated expected start time of the node. The job flow management apparatus according to claim 1, wherein: Job attribute data indicating the attribute of each node constituting the job flow represented by each node indicating a job or a junction is stored in advance, and each job is transferred to a predetermined execution device and executed. And a job flow management device that receives the execution result from the execution device,
As the job attribute data, a time constraint indicating a time at which the job should be executed, a target execution time indicating a target of the time at which the job should be executed, and an expected time that is expected to be executed by the job Pre-store data including execution time,
According to a schedule preset in the job attribute data, the schedule management means instructs the execution start of the job flow,
Based on the job attribute data, the execution order determination unit determines the execution order in the order of execution that is the latest among the jobs that join the specific joining point in the job flow,
Using the time constraint and the expected execution time while tracing each node from the end point of the job flow, the expected start time calculation unit calculates the expected start time of the job sequentially from the job whose execution order is late,
When the calculated expected start time is delayed by a predetermined time or more compared to a target execution time set in advance in the job attribute data, a warning transmission unit transmits a warning to the user,
A job flow execution start time prediction method characterized in that a job input unit transfers each job to the execution device, requests execution, and receives the execution result from the execution device. A time constraint indicating the attribute of each of the nodes constituting the job flow represented by each node indicating a job or a confluence, and indicating the time when the job should be executed, and the target time when the job should be executed Job attribute data including a target execution time and an expected execution time that is expected to be executed for the job are stored in advance, and each job is transferred to a preset execution device. In the job flow management apparatus that requests execution and receives the execution result from the execution apparatus,
A procedure for instructing execution start of the job flow according to a schedule set in advance in the job attribute data,
A procedure for determining an execution order in the order of the latest execution among each job that merges with a specific merge point in the job flow based on the job attribute data;
A procedure for calculating the expected start time of the job sequentially from the job with the lower execution order by using the time constraint and the expected execution time while tracing each node from the end point of the job flow.
A procedure for issuing a warning to the user when the calculated expected start time is delayed by a predetermined time or more compared to a target execution time preset in the job attribute data;
And a job flow execution start time prediction program for causing a computer to execute a procedure of transferring each job to the execution device and requesting execution thereof and receiving the execution result from the execution device.

Images