Query performance on WebLogic...

swhitmire

Reputable Poster
We're in the process of migrating from WebSphere to WebLogic, and we're having a strange problem with query performance. An initial find in a screen will take two seconds or less, and repeating that find in the next few seconds is even quicker. But repeating the find at some later time takes significantly longer, usually around six seconds. Initially we thought this was happening after 30 seconds, but now that we've tested with three different servers, the time at which this happens seems to vary -- sometimes it'll be OK two minutes later or more, but sometimes it'll be bad after just like 25 seconds. The delay also seems to happen sometimes when switching between open applications. If it was simply that after a time the query/application switch speed was the same as the first time, I wouldn't be concerned, but it's consistently 4-5 seconds more, which I think is going to be frustrating for users. Anyone have any ideas? This is WebSphere 12.1.3.0.0 on RedHat 6.8.
 
Scotti, We are also in the process of migrating our WebSphere servers to WebLogic 12.1.3 but on Windows 2012. It would be helpful if you can specify what is memory size of the JVM's, including what JDK version has been used. So far in our testing in PY with a concurrent user load of 60-70 a JVM running on 1.8.60 with 4 GB of memory has worked well for us with no performance issues. We are on 9.1 (with 9.1.5.5) in the process of migrating to TR 9.2.0.5.

I have also experimented with the G1 Garbage collection algorithm and found it be making a difference.

For a deeper troubleshooting it could help if you try and perform a profiling of your Garbage Collection.

Thanks,
Soumen
 
I have 4GB assigned to the JVM, with JDK 1.7.0_79. The overall performance seems fine, except for this specific problem. We're just doing our initial testing -- this happens even with only one user on and one screen open, so I doubt it's a problem with garbage collection. It seems like it must be something about some kind of timeout or caching, but my research so far hasn't found anything that seems to make a difference, and I just can't understand why the subsequent finds are so much slower than even the initial one. I've had a case open with Oracle for a while and they haven't come up with a good answer. I wondered if it could be a browser issue, but it happens in all of them -- IE, Chrome, and Firefox.
 
Have you verified cache settings, webtime run interactivity--number of browsers each user can open................also JVM max size may be too high, have you tried 2GB?
 
I've now got six WebLogic servers set up, and they all have this behavior, even if only one user is logged in. Four of them have 4096/4096 for the JVM, one has 2048/2048, and one has 32/1024, so it doesn't seem that the size is making a difference.
 
Back
Top