D
dbeavon
Guest
I'm stumped. It seems like you've narrowed down the scope of the problem to something that is entirely a CPU bottleneck. If you watch your system resources, can you verify that this is running with 100% CPU utilization (or at least it uses the single-threaded equivalent of 100% CPU)? If not, then there are other unknown factors and resources that may be participating in your repro. Also is there only * one* PASOE session that is active in the agent while you are running your repro? If there are any other sessions that are active within the agent process then there might be concurrency-related bottlenecks that are slowing things down, (eg I'm thinking of things like the artificial bottleneck that prevents more than one thread in a process from using XML data at a given moment of time ... see PERFORMANCE OF CONCURRENT READ-XML CALLS ON PASOE SLOWER - Forum - OpenEdge General - Progress Community ) Or perhaps your new PASOE server has a virtualization problem that might be short-changing you on CPU. Maybe you should run both versions of your tests on the same hardware - this will take the hardware and virtualization out of the equation. It would seem odd to me if PASOE was the only thing that was responsible for the slowdown. I'd guess the upgrade in OpenEdge versions would be more likely to cause this performance regression. The hosting of the code into a PASOE msagent process should not normally make any massive difference.
Continue reading...
Continue reading...