We have just moved to new Intelliflex 16.20 and we are facing performance degradation on system.
Strange thing is that we see like 2-3 queries running on system but still we get AWT exhaustion alerts.
Only thing for performance degradation that have been identifed now is that "TIM" has been enabled and it used lot of I/O.
We are planning to disable TIM.
So, how can high I/O can cause AWT exhaustion on system. I mean is there any relation between them ?
TIM on its own doesn't exactly "use I/O"; rather some I/O operations will be handled as in-memory operations. It seems unlikely that disabling TIM will solve this issue.
AMP Worker Tasks do all the work for query steps, including I/O requests, so yes - there is a relationship.
Did you perhaps inadvertently change the number of AWTs per AMP on the new platform? Or the number of AWTs reserved for expedited work? When do you see AWT exhaustion? All the time or only when certain jobs/queries are running? At irregular intervals, or fairly regular times (e.g. every 10 minutes)?
Reducing TIM value did helped system performance at that time but i guess you are right, we are still facing high performace degradation, massive flow controls and AWTs getting exhausted, Database restarting and becoming unavailable too.
AWTs limit is the same as per old system. There is no specific pattern or intervals observed so far for Flow Control. It comes over and over and at no specified times.