Hello,
We have migrated a large J2EE application from Weblogic to Wildfly. It heavily uses EJB and JMS. After changes on certain code specific to Wildfly, we started doing performance testing on Wildfly and found below statistics:
OS: Linux Centos (AWS Ec2)
CPU: 8 VCPU
Memory: 64 GB
JDK: Oracle JDK 1.8
JAVA Heap: 10g (min and max)
Response in Milliseconds for specific use case.
Wildfly Weblogic
Min 26 18
Max 486 70
AVG. 36 23
90 P 37 26
TPS: 105 144. for 5 threads running for 10 mins.
We have tried following tuning options:
- OS Tuning – Ulimit, Hugepages
- JVM settings – changed GC strategy, heap, metaspace and others
- WildFly Container – JMS lookup approach, logs, thread pooling optimization, Data source tuning, IO, EJB and others
- Application refactoring – optimized EJB and JMS invocation, pass by ref and others
After all the above changes, unable to improve the TPS from 105.
is there any performance benchmark available between Weblogic and Wildfly?
is there any degradation observed when moving to open source Wildfly from Weblogic?
Anybody has any experience which can share on this, will be helpful.
Thanks,
Pankaj K.