for getting a better performance we switched our Webservice from axis1 (1.4=
) to axis2(1.5 also tried 1.4.1). As of this change we also switched parts = of our database schema, and so we dind=B4t test performance until now. I wr= ote a DB transformer that created the new database out of the old one. So w= e had valid and a hughe amount of data. I did some tests, and discovered th= at a request that took about 1second on axis1.4 now takes up to 10 seconds =
- (the processing of my own code takes about 300ms). I tuned all logging st= uff on and discovered that the xml generation takes ALOT of time. The retur= ned object holds a list of about 3000 objects with each holding some string= s and 2 other objects holding 2strings and 1 int.
I serialized the Whole response to get an idea how big the response is. Thi= s way it say it=B4s 300kb.
The server and the client run on the same machine (faster) or on different = machines (slightly slower, still a 1gbit connection). So the 300kb should b= e no problem - even if the XML overhead is 5 time as big as the data...
Funny is, that Axis2 1.5 takes like 3 times longer than Axis2 1.4.1 does. B= ut both performances are not really acceptable. Especially because axis2 is= MEANT to be faster when it comes to bigger datasets.
Did anyone else have these problems and can help me out?
Some info that might be of any interest:
Running 5 services simultanously, where all oft hem use the same libaries o= f different versions (or the same). (Axis2 webpage sais this should be fine=
Running tomcat version 6.0
No additional Modules installed.
No other servlets running.
It=B4s not that axis is WATING when i request some lists, i can see the cpu= working in taskmanager.
I reinstalled tomcat
I redeployed axis2 1.5/1.4.1
I tried to work with 1 service only.
I had a look if MTOM is turned on.
I tried with java and C# client.
If i forgot to mention something important, please ask me!