Hi,
I did some 'benchmarking' of webservices provided by our OpenEdge 10.2B application. My tests were done in a developement enviroment so hopefully we get 2 to 3 times better performance in production.
I get an average of about .2 seconds per call when calling the service on the same machine and 3 to 4 times that when calling it from another machine.
What would be the estimated gain when we would use 11.5 with the pacific appserver? Would is be as significant as the 700+% that was reported doing benchmarks with AIA vs. Pacific Appserver?
Richard
[/collapse]Thread created by Richard.KeltersHi,
I did some 'benchmarking' of webservices provided by our OpenEdge 10.2B application. My tests were done in a developement enviroment so hopefully we get 2 to 3 times better performance in production.
I get an average of about .2 seconds per call when calling the service on the same machine and 3 to 4 times that when calling it from another machine.
What would be the estimated gain when we would use 11.5 with the pacific appserver? Would is be as significant as the 700+% that was reported doing benchmarks with AIA vs. Pacific Appserver?
Richard
Stop receiving emails on this subject.Flag this post as spam/abuse.
[/collapse]Thread created by Richard.KeltersHi,
I did some 'benchmarking' of webservices provided by our OpenEdge 10.2B application. My tests were done in a developement enviroment so hopefully we get 2 to 3 times better performance in production.
I get an average of about .2 seconds per call when calling the service on the same machine and 3 to 4 times that when calling it from another machine.
What would be the estimated gain when we would use 11.5 with the pacific appserver? Would is be as significant as the 700+% that was reported doing benchmarks with AIA vs. Pacific Appserver?
Richard
Stop receiving emails on this subject.Flag this post as spam/abuse.
Hi Richard,
A non-academic test of the REST adapter showed a simple ping call in the classic typically took 40 ms, whereas the same functionality took about 13 ms. Having said that, when loading testing (with SOAP UI, 5 threads) that ping operation I got an average call time of 3 ms. Since I didn't test that scenario in the classical AppServer I can't really tell what the 3 ms is really worth. It is however save to say that the per call overhead (latency) is drastically cut back. What this is worth depends on the scenario you're running (are you running a high volume of short calls, or a few with long processing times). In short, PASOE seems to reduce the latency more than the actual (time wise) performance of the process.
In relation to what Mike said: for SOAP to show those numbers we have to wait a little longer. On the other hand, I'm working with SOAP web services at one of our customers and they feel pretty fast too, even without the 11.5.1 update. Is 11.5.1 still scheduled for June?
PS. if your setup adds 600 ms to a networked call I would do some more checks, something is not performing as it should.
Thanks you all for replying.
We're going to upgrade to 11.5, I'll inform our client to hold it until 11.5.1 is out in order to profit from the bugfix.
As our tests results are way too slow and we need dramatic improvements, we're going to look into other solutions. One of my tests was 300 synchronous small calls, we can speed that up by providing the same information in 1 call.
Yes we know our client network has some issues (@Bronco you know who it is) but I rather not spend my time and energy in management/politics.