Hello. We are experiencing interesting flow behavior related to setting
Concurrent Tasks for an ExecuteScript processor (ES v1.11.4). In an effort
to improve flow throughput we increase Concurrent Tasks on the processor.
As we increase to 5, we do indeed see increase in throughput. But beyond
that, if we increase Concurrent Tasks the throughput begins to fall off.

We subsequently tried a configuration where we distribute to two
like-configured ES processors, each set to 5 Concurrent Tasks. Throughput
is better than when we run as described above.

Why does it appear to be better to distribute to two processors each
running 5 Concurrent Tasks than it is to crank up Concurrent Tasks on a
single processor to 10? Do these situations differ in JVM memory
consumption? Does the first case impact Core Load Average more than the
latter case?

Thank you in advance for any insights.

Reply via email to