As it turned out the HC executor wastes 50ms by using a hardcoded Thread.sleep() which does not sound like a lot (and often it's not a problem), but some use cases it's not ideal (e.g. when calling it from a load balancer). The pull request in SLING-5874 fixes it... Carsten or Betrand, the two of you were involved in this particular piece of code (the very line with the sleep came from myself, blaming myself ;-)), maybe you could have a look and commit it? I tested it with JMeter, there should not be any problems with it.

Regards
Georg

Reply via email to