Managing bottlenecks

A bottleneck occurs when concurrent requests are collecting in a concurrent manager's run queue. At some peak periods, the concurrent manager does not process the requests fast enough, and a greater number of requests collect in the concurrent manager's queue. In addition, these pending requests can wait for a concurrent manager long enough to cross the long-pending threshold. Bottlenecks are usually discovered when waiting time statistics increase and concurrent requests begin crossing the long-pending threshold with an increasing frequency.

If the excessive waiting times occur during predictable time periods or due to specific requests, you can set up additional concurrent managers or target processes and assign them to a specific work shift. You can also assign certain requests to specific concurrent managers so they do not interfere with other requests.

Before you identify bottlenecks and peak waiting times, you should have enough history to study trends and patterns in your environment.

This section contains information about the following tasks:

Related topics

Monitoring-concurrent-managers
Monitoring-concurrent-requests