JP4456465B2 - Load stabilization method for automatic control system - Google Patents

Description

The present invention relates to a load stabilization method for an automatic control system that centrally manages the operations of a plurality of distributed systems.
In a fully automated system that operates continuously all day long, the start operation and end operation are automatically controlled while monitoring the status of a plurality of distributed systems with a central management system. In such a fully automatic system, the load is stabilized by a time unit activation method in which each distributed system is activated at regular time intervals or an event drive method in which each distributed system is activated every time the state of each distributed system changes. Then, it is necessary to further develop such a load stabilization function to prevent a reduction in processing efficiency in the central management system due to load concentration.

  FIG. 13 shows an example of a fully automated system. A number of distributed systems 2 are connected to the central management system 1, and the start and end operations of each distributed system 2 are automatically controlled by the central management system 1 for continuous all day operation based on the start and end conditions of each distributed system. Is done.

  FIG. 14 shows a semiconductor manufacturing data conversion processing system as an example of a fully automated system. The systems A to D are distributed systems, and the case where the system D is automatically started by the central management system 1 based on the information of the systems A to C is shown. For example, system A is an order management system, system B is an area management system, system C is a manufacturing parameter management system, and system D is a system that creates manufacturing data based on design data and information of systems A to C. It is.

  When the central management system 1 has all the information necessary for the systems A to C, it outputs a system startup JOB queue to the system D, and the system D generates manufacturing data from the design data based on the JOB queue. To start.

  FIG. 15 shows the operation of the conventional time unit activation method in the fully automated system shown in FIG. The central management system 1 inquires of the systems A to C at regular intervals whether or not necessary information is prepared in the systems A to C (steps 1 to 3), and takes in the information when the necessary information is prepared ( Step 4), a direct job queue for starting the system D is generated and stored (step 5).

  FIG. 16 shows the operation of a conventional event-driven system in the fully automated system shown in FIG. The central management system 1 monitors whether or not an event (state change) has occurred in the systems A to C (step 11). When the occurrence of the event is received, whether or not necessary information has been prepared in the systems A to C. Are inquired of the systems A to C (step 12). Then, when necessary information is prepared in the systems A to C, the information is taken in (step 14), and a direct job queue for starting the system D is generated and stored (step 15).

In any method, it is desirable to reduce the load on the central management system and each distributed system without impairing the responsiveness of the central management system depending on the situation of each distributed system.
Patent Document 1 discloses a database system similar to the fully automated system as described above.
JP 2001-14336 A

  In a system that operates by the time unit activation method as described above, whether or not necessary information is prepared in the systems A to C is determined by the systems A to C at regular intervals regardless of whether or not an event has occurred in the distributed system. In order to make an inquiry, the central management system 1 increases the number of processing steps for the inquiry.

  That is, in order to improve the responsiveness from the preparation of necessary information in the systems A to C to the generation of the JOB queue directly, it is necessary to shorten the time interval for making an inquiry. Processing man-hours will increase. On the other hand, if the time interval for making an inquiry is lengthened, the responsiveness decreases. Further, it is necessary to continuously detect whether or not an event has occurred in the systems A to C.

Therefore, there is a problem that the load on the central management system 1 increases if sufficient responsiveness is to be ensured.
On the other hand, in the event driving method, since an inquiry is made only when an event occurs in the systems A to C, there is no useless inquiry by the central management system 1, and an increase in load is suppressed.

  In addition, since a JOB queue is directly generated in response to the occurrence of an event, the response is superior to the time unit activation method, and information is replaced in the systems A to C after the generation of the direct JOB queue. However, it is also possible to directly generate a new JOB queue based on the occurrence of the new event.

  However, since the number of inquiries and the number of direct job queue generations increase according to the occurrence of an event, if the occurrence of an event for automatically starting the system D increases, the load on the central management system 1 increases. There is.

  An object of the present invention is to provide an automatic control system having excellent responsiveness by suppressing an increase in load even if the occurrence of an event increases.

The above object is a “direct JOB queue” which is a JOB queue for starting a controlled distributed system corresponding to an event based on the occurrence of any event of a plurality of distributed systems and the load state of the central management system. Or an automatic control system that generates a “state change recording queue”, which is a JOB queue that records only state changes , in accordance with the load on the central management system and the distributed system on the controlled side. The storage time for storing the “state change record queue” is adjusted, and the state change record that becomes unnecessary with the passage of time during the storage time is deleted from the “ state change record queue ” , and the state change after deletion This is achieved by a load stabilization method of an automatic control system that generates the “ direct JOB queue ” based on a recording queue.

  According to the present invention, even if the occurrence of an event increases, it is possible to provide an automatic control system with excellent responsiveness by suppressing an increase in load.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings. This embodiment shows a system similar to the semiconductor manufacturing data conversion processing system shown in FIG. 14 and is the same as the configuration shown in FIG. 14 except for the function of the central management system 11.

  The central management system 11 shown in FIG. 1 basically operates in an event-driven manner, and similarly to the conventional example, a function for directly generating a JOB queue based on the occurrence of an event in the systems A to C, and the systems A to C And a function for generating a state change recording queue based on the occurrence of an event at, and a function for erasing unnecessary state change recording.

  The central management system 11 generates a state change recording queue (including a state change record as a JOB) based on the occurrence of an event by using a state change database search function provided in advance. Hereinafter, the function of generating the state change recording queue of the central management system 11 will be described.

  FIG. 2 shows an example of a basic operation for generating a state change recording queue based on a plurality of information of the systems A to C. The items A1 to A3 of the system A, the items B1 and B2 of the system B, and the system C A case where a JOB queue is generated based on the C1 to C4 items in FIG.

  First, when information on each item is prepared in the systems A to C, the information is sequentially taken in and held in the central management system 11. At this time, since the information of each item of the systems A to C is information necessary for starting the system D, the items A1 to A3, the items B1 and B2, and the items C1 to C4 are all set as a common LA group. .

  Then, as shown in FIG. 3, the three states that the LA group has a state change based on the fact that necessary information is prepared in the systems A to C and the event has changed once in each of the systems A to C. Change recording queues (1) to (3) are generated.

  Next, for example, when the item A2 of the system A is changed after 10 minutes, a fourth state change recording queue (4) in which there is a state change in the LA group is generated. Further, when the item B1 of the system B is further changed after 50 minutes, a fifth state change recording queue (5) in which there is a state change in the LA group is generated. Further, when the item C4 of the system C is further changed after 30 minutes, a sixth state change recording queue (6) in which there is a state change in the LA group is generated.

  Here, if a JOB queue is directly generated for each state change recording queue (1) to (6), it becomes the same as a normal event driving method, but in this embodiment, a load environment parameter described later is used. Based on this, as the load on the system D increases, the number of direct job queue generations is reduced in order to reduce the load on the central management system 11. That is, if the job queue is directly generated based on the last state change recording queue (6) 90 minutes after the state change recording queues (1) to (3) are generated, the system D uses the latest information. This is because required manufacturing data can be generated.

However, in the normal case, since the end of the change cannot be predicted, it is difficult to determine the last state change.
Therefore, the state change recording queue storage time is adjusted according to the load, and the JOB queue is directly generated based on the last state change recording queue required during the storage time.

  FIG. 4 shows a state change recording queue generating operation when manufacturing data for a plurality of products is generated based on the information of the systems A to C and the system E. In this case, three types of manufacturing data are generated by the system D based on the information of the three groups LA to LC from the systems A to C. The information of the system E is information commonly used for the groups LA to LC.

  FIG. 5 shows the state change recording queue when the information of the system E is changed three times in such a case. As shown in the drawing, three state change recording queues are generated for each of the groups LA to LC by three changes of the information of the system E.

  In this case, production data corresponding to each of the groups LA to LC can be generated by the system D if the job queue is directly generated based on at least the state change recording queues (7) to (9) according to the load environment parameters described later. It is. Of course, the JOB queue can be directly generated nine times based on the state change recording queues (1) to (9).

  FIG. 6 shows a case where the groups LA to LC are defined as one order 0001 group in the operation shown in FIG. If defined in this way, each group LA to LC is recognized as one group, and therefore, nine state change recording queues (1) to (9) when the groups LA to LC are made into independent groups are included in the system. E is collected in three state change recording queues (11) to (13) due to three changes of information.

Therefore, the load of the central management system 11 is reduced by directly generating the JOB queue three times by changing the information of the system E three times.
Of course, depending on the load, the JOB queue can be directly generated only once based on the last state change recording queue (13).

  However, in such a case, for example, even when a state change occurs only in the group LA, a state change recording queue is generated for all the groups LA to LC, and a JOB queue is directly generated. It is preferable to provide a later-described pass filter function.

  FIG. 7 shows the operation of the filter function of the central management system 11. This filter function sets a flag indicating whether or not a JOB queue has been generated directly based on the information of the groups LA to LC when the order 0001 group as described above is defined. The group in which the flag is set intends to reduce the load by preventing the system D from being activated based on the information of the group even if a state change occurs in another group.

  That is, a state change recording queue is generated based on the occurrence of an event (steps 21 and 22). Next, it is determined whether or not the information of the system E is changed (step 23). If the information of the system E is changed, the activated flag is set in the groups LA to LC (step 24). Then, a direct job queue for starting the system D is generated (step 25).

  In step 23, it is determined whether or not the activated flag has already been set for the groups LA to LC instead of changing the information of the system E (step 26). If no flag is set, the process proceeds to step 24.

  In step 26, the activated flag is already set for the groups LA to LC. For example, when only the information of the system A is changed, the activated flag is canceled for the group LA (step 27). Based on the change information of the group LA, a direct job queue for starting the system D is generated (step 28), and no direct job queue is generated for the groups LB and LC.

Then, an activated flag is set for the group LA (step 29), and the operation is terminated.
FIG. 8 shows a job queue distribution function based on load environment parameters. This function is a function for efficiently controlling each distributed system by adjusting the number of direct job queues and state change recording queues stored in the central management system 11.

  That is, based on the occurrence of an event, the job status stored directly in the job queue is acquired (steps 31 and 32), and the load environment parameter corresponding to the external environment is automatically corrected (step 33).

  The load environment parameters are the upper limit value of the direct JOB queue, the setting period of the upper limit value, the storage state of the direct JOB queue, etc. In step 33, the upper limit value of the direct JOB queue is set to 5, for example, and the setting period is 90. Day. In step 33, if there is a timing when the number of direct JOB queue storage (unprocessed number) becomes 0 even though the state change recording queue is stored, the upper limit value of the direct JOB queue is reset to 6. Set.

On the other hand, if the upper limit value of the direct JOB queue is set to 6 and the state that is not changed as it is continues for 90 days or more, the upper limit value of the direct JOB queue is reset to 5.
Based on such load environment parameters, the storage state of the direct job queue is monitored, and when the number of direct job queue storage is large, that is, when the load is high, a state change recording queue is generated, and when the load is low, the direct job queue is generated. Is generated (step 34).

  Next, the operation of the central management system having the above functions will be described. FIG. 9 shows a basic operation of the central management system 11 for starting each system shown in FIG. That is, when the start condition for starting the distributed system, that is, necessary information is prepared, the central management system 11 directly stores the start JOB as a JOB queue (step 41). At this time, the storage status of the JOB queue is directly notified to the load environment parameter.

  Next, when the generation of the direct JOB queue is completed, it is determined whether or not the processing operation of the terminated system affects the event driving of other distributed systems, that is, whether or not there is a direct JOB queue remaining (step 42). . If there is no influence, the system waits until the start conditions for starting the next distributed system are met.

  In step 32, if there is a direct job queue remaining but its number is small, an activation job for the next distributed system is generated as a direct job queue. If the number of remaining direct job queues is large, a state change recording queue is generated (step 43).

Then, the state change record queue is stored, and the state change record which becomes unnecessary during the storage period is deleted (step 44).
Next, when the number of unprocessed direct JOB queues decreases according to the state of the load environment parameter, a direct JOB queue corresponding to the state change recording queue is generated. At this time, unnecessary state change records are deleted by the filter function, and a JOB queue is directly generated (step 45).

10 to 12 show specific operations in the semiconductor manufacturing data conversion processing system shown in FIG.
FIG. 10 shows a case where the load on the system D is low. In this case, for the six events (1) to (6), the central management system 11 makes six inquiries to the systems A to C, and necessary information is prepared (3) to ( For the four events in 6), a JOB queue is directly generated and the system D is activated four times.

  FIG. 11 shows a case where the load on the system D is high. In this case, for the six events (1) to (6), the central management system 11 generates and stores a state change recording queue for each event, and the state change recording queue only for the last event (6). The system A to C is inquired based on the above, a job queue is directly generated, and the system D is activated once.

  FIG. 12 shows a case where the load on the system D changes to a high state or a low state. That is, when the load on the system D is high due to the events (1) and (2), the central management system 11 generates and stores a state change record queue. When the load on the system D is low due to the events (3) and (4), the central management system 11 makes an inquiry, generates a JOB queue directly, and starts the system D. Further, when the load on the system D is high in the events (5) and (6), the central management system 11 generates and stores the state change record queues respectively, and only in the state change record queue for the last event (6). Based on this, the system A to C are inquired, a job queue is directly generated, and the system D is activated.

  Therefore, in the case of FIGS. 11 and 12 in which the load on the system D is high, the number of inquiries of the central management system 11 and the number of activations of the system D are reduced compared to the case in which the load shown in FIG. 10 is low.

In the automatic control system as described above, the following operational effects can be obtained.
(1) A state change recording queue is generated based on the occurrence of an event, and unnecessary ones are deleted from the state change recording queue as time elapses, and a direct job is executed based on the state change recording queue after the deletion. Since the queue is generated, an increase in load due to an increase in events can be suppressed.
(2) Since the system can be activated based on the occurrence of an event as in the conventional event driving method, responsiveness to the occurrence of the event can be ensured.
(3) By grouping the state change recording queues generated by the occurrence of an event, it is possible to suppress an increase in load accompanying an increase in the event.
(4) When the job queue is directly generated based on the state change recording queue, it is possible to prevent the system from being duplicated due to the grouping of the state change recording queue by the filter function.
(5) Since it is possible to select whether to generate a state change recording queue or to directly generate a JOB queue when an event occurs based on a load environment parameter, a load stabilization function according to the load state is provided. be able to.

The above embodiment may be modified as follows.
-You may implement similarly to fully automatic systems other than the semiconductor manufacturing data conversion processing system.

It is a block diagram which shows the manufacturing data conversion processing system of a semiconductor. It is explanatory drawing which shows operation | movement of a central management system. It is explanatory drawing which shows a state change recording queue. It is explanatory drawing which shows the production | generation operation | movement of a state change recording queue. It is explanatory drawing which shows a state change recording queue. It is explanatory drawing which shows a state change recording queue. It is a flowchart which shows operation | movement of a filter function. It is a flowchart which shows distribution operation | movement of the JOB queue based on a load environment parameter. It is a flowchart which shows the basic operation | movement of a central management system. It is explanatory drawing which shows the specific operation | movement of a manufacturing data conversion processing system. It is explanatory drawing which shows the specific operation | movement of a manufacturing data conversion processing system. It is explanatory drawing which shows the specific operation | movement of a manufacturing data conversion processing system. 1 is a block diagram illustrating a fully automated system. It is a block diagram which shows the conventional manufacturing data conversion processing system. It is a flowchart which shows the operation | movement of the conventional time unit drive system. It is a flowchart which shows the operation | movement of the conventional event drive system.

Explanation of symbols

2 Distributed system 11 Central management system

Claims (4)

A “direct JOB queue” or status change that is a JOB queue that activates the controlled distributed system corresponding to the event based on the occurrence of any event of a plurality of distributed systems and the load state of the central management system An automatic control system that generates a “state change recording queue” that is a JOB queue in which only the data is recorded by the central management system ,
Wherein adjusting the dwell time to store the "state change recording queue" in accordance with the load of the central management system and the controlled side of the distributed system, wherein from the "state change recording queue", the time during said dwell time A method for stabilizing the load of an automatic control system, wherein state change records that become unnecessary as time elapses are deleted, and the " direct JOB queue " is generated based on a state change record queue after deletion. The grouping of "state-change recording queue", the load stabilizing method of an automatic control system of claim 1 Symbol placement and generates the "direct JOB queue" for each such group. 3. The activated flag is set for each event corresponding to each grouped “ state change recording queue ” , and the redundant activation of the distributed system is prevented based on the activated flag. Load stabilization method for automatic control system. Load states of the central management system and the distributed system on the controlled side are stored in load environment parameters, and based on the load environment parameters, the “ state change recording queue ” and the “ direct JOB queue ” are generated by the occurrence of the event. The load stabilization method for an automatic control system according to any one of claims 1 to 3 , wherein which one of the two is generated is distributed.

Images