8.3.2 Procedure for requesting the maximum number of concurrent executions and pending queue
You can tune the maximum number of concurrent executions and pending queue size according to the following procedure:
-
Increase the multiplicity of requests by using tools such as the load generation tool.
At this point, if the CPU usage on the server-side is 80% to 90% proceed to step 2. If the CPU usage is not 80% to 90%, the performance is low and it is not possible to increase the throughput, it is likely that there are bottlenecks in the I/O processing and exclusion processing. In such case, identify the process with the bottleneck and enhance the performance.
-
Set the multiplicity in which the CPU usage on the server-side has reached 80% to 90% as the maximum number of concurrent executions in the tuning parameters.
The throughput at this stage is the maximum throughput of a single machine. Enhance the hardware if you require a higher throughput.
-
Check whether maximum throughput can be maintained by applying additional load by using tools such as load generation tool.
If a maximum throughput cannot be maintained, revise the tuning parameter to prevent loads that exceed the maximum throughput.
-
Estimate the number of requests when the actual system is temporarily overloaded and when the loading is at a peak state and decide the size of the pending queue.
-
In an application having a hierarchical structure, revise the tuning parameter so that the size of the pending queue and number of concurrent executions in each layer can be balanced.