WM Performance glitch

I am performance testing some integration components and have come across some strange results. We built a broker client that does a variable number of request/reply transactions and records the elapsed time for each. Most of the transactions take fractions of a second but some of them take just over 5 seconds. In a situation where five clients simultaneously publish 1000 of the same request/reply event (same eventtype, same data) we will have about 4960 events return in approximately .01 seconds; the rest return in approximately 5.01 seconds. We have run the request/reply test against an application integration component, an Oracle DB adapter component, and a stub ILA component and we consistently get the same results. We have gotten similar results publishing these events from apps and the EventTracker (not just our Broker Client). Any thoughts would be appreciated.

Did you get this one solved yet, Brian? If not…

It seems strange that 4960 of them occur in .01 seconds and the last 40 take 5.01 seconds. That same “.01” is what makes me wonder.

What is the total transaction time for all 5000 events? Is it (5000 events * .01 seconds)?

Do the 5+ seconds transactions all occur in a series or are they mixed throughout?