OpenTP1 Version 7 Operation
If the service requests from clients start to be delayed in the service processing stage, they may stay too long in the schedule queue because they cannot be fetched.
Therefore, OpenTP1 checks (monitors) the service requests remaining in the schedule queue at a specified interval for each user server. This function is called schedule queue congestion monitoring, and is available only for user servers (SPPs). The following figure shows an example of service requests remaining in the schedule queue.
Figure 3-3 Service requests remaining in the schedule queue
If the number of service requests remaining in the schedule queue exceeds the value specified for the applicable operand in the system definition during schedule queue congestion monitoring, OpenTP1 outputs message KFCA00833-W. Depending on the specification of the operand, OpenTP1 system also outputs message KFCA00834-E and goes down (forced termination).
To use schedule queue congestion monitoring, specify the following operands in the user service definition or the user service default definition. For details about each operand, see the manual OpenTP1 System Definition.
You can specify the above operands only on SPP user servers. If you specify these operands on a RAP-processing server or an MHP server, schedule queue congestion monitoring will not be enabled. If you omit the stay_watch_queue_count operand or specify 0 for it, the values specified for operands 2 to 5 have no effect.
This subsection describes the processing sequence of schedule queue congestion monitoring. Note that the user server (SPP) has already been started.
This subsection describes an example of schedule queue congestion monitoring when the following values are specified for the operands in the user service definition:
Figure 3-4 Example of schedule queue congestion monitoring
In Figure 3-4, judgment of whether the schedule queue is congested occurs in the period from C2 to C5 and in the period starting at C8. Judgment starts because of the specification stay watch_queue_count=30. During each period, judgment of queue congestion is performed at the interval specified in the stay watch_check_interval operand. At this interval, OpenTP1 judges whether the schedule queue is congested according to the expression for judging congestion, which is based on the number of service requests that are processed and the processing rate of service requests.
Expression for judging congestion of the schedule queue:
The expression for the example in Figure 3-4 is as follows:
The following table shows the number of service requests that are processed in Figure 3-4, and the results of evaluating the expression for judging congestion of the schedule queue.
n | Number of processed service requests (Pn-1 - Bn) |
Number of service requests expected to be processed (m1 x Pn-1) |
Result of evaluating the expression for judging congestion of the schedule queue (Pn-1-Bn < m1 x Pn-1) |
---|---|---|---|
0 | -- | -- | Not applicable. |
1 | 18-9=9 | 0.7 x 18=12.6 | Not applicable. |
2 | 28-25=3 | 0.7 x 28=19.2 | Judgment of whether the schedule queue is congested starts. |
3 | 32-8=24 | 0.7 x 32=22.4 | OpenTP1 continues online operation. |
4 | 45-13=32 | 0.7 x 45=31.5 | OpenTP1 continues online operation. |
5 | 35-0=35 | 0.7 x 35=24.5 | OpenTP1 continues online operation. |
6 | 30-3=27 | 0.7 x 30=21 | Not applicable. |
7 | 11-5=6 | 0.7 x 11=7.7 | Not applicable. |
8 | 17-15=2 | 0.7 x 17=11.9 | Judgment of whether the schedule queue is congested starts. |
9 | 32-29=3 | 0.7 x 32=22.4 | The OpenTP1 system goes down. |
When n equals 1 or 7 in the above table, the expression for judging congestion of the schedule queue evaluates to true. However, the OpenTP1 system does not go down because judgment of whether the schedule queue is congested is not being performed. When n equals 2 or 8, the OpenTP1 system does not go down because the judgment process is only beginning.
When n equals 9, the OpenTP1 system goes down because judgment of whether the schedule queue is congested is being performed, the expression for judging congestion of the schedule queue evaluates to true, and stay_watch_abort=Y is specified.
All Rights Reserved. Copyright (C) 2006, 2010, Hitachi, Ltd.