3.1.5 Event monitoring performance
If you intend to monitor events using JP1/AJS3 event jobs and jobnets with start conditions, you must first estimate the number of event jobs that will be executed and the number of generated events, and keep monitoring to within the JP1/AJS3 processing performance.
This subsection describes how to estimate and set the processing performance of JP1/AJS3 in regard to monitoring events.
Processing performance estimates for event monitoring are based on how long it takes an event job or a jobnet with a start condition to be ready to process the next event after a preceding event occurs. If events occur at shorter intervals than the estimated length of time, a bottleneck of event jobs and jobnets with start conditions might occur, potentially delaying the processing of other jobnets. In cases where you can predict how frequently events will be generated, we recommend setting monitoring conditions that will not result in events occurring at shorter intervals than predicted.
Bottlenecks might also occur when a number of events are being monitored and two or more occur at roughly the same time. Avoid monitoring types of events that are likely to occur at the same time.
The processing time for monitoring one event is approximately 0.6 seconds, depending on the hardware and other factors.
You can also estimate the processing capacity for monitoring events by a scheduler service, based on how many events are generated at the scheduler service per unit time. We recommend a system design that will keep the number of events reported to the scheduler service at a level well below the value calculated using the following formula:
- At jobnet registration
-
Length of time until the jobnet can be executed (seconds) = number of event jobs# x 0.8
- #
-
Total number of event jobs (including jobnets with start conditions) registered for execution at the same time
- At event occurrence
-
Number of events that can be processed = unit of time (seconds) / 0.6
- Example: Calculate the number of events that can be processed in 60 seconds.
-
Number of events that can be processed = 60 / 0.6 = 100
Thus, 100 jobs can be processed in 60 seconds.
This formula gives the peak processing performance (that is, the processing capacity when only event jobs and jobnets with start conditions are processed).
Keep the number of events to no more than 1,000 per hour per scheduler service, even at peak times.
- Performance when the event order option is used
-
When the event order option is used, an agent that has sent an event to the manager does not send the next event until it receives a reply from the manager. Processing that checks for a reply from the manager is performed at regular intervals (every ten seconds), and the agent sends the next event only after a reply is received from the manager. Accordingly, the maximum number of events that can be processed per hour is 360. We recommend that you estimate 180 events being processed per hour when the event order option is specified, although the actual value will depend on how often events occur in the particular system.