Internals of a stateful JPD and impact of having 1000's of them - Weblogic
This is a discussion on Internals of a stateful JPD and impact of having 1000's of them - Weblogic ; Hi
If we have a scenario in which we have literally thousands of stateful JPD =
instances waiting on say different events that would then make them "runnin=
g", what would be the impact performance wise on the container? (this ...
-
Internals of a stateful JPD and impact of having 1000's of them
Hi
If we have a scenario in which we have literally thousands of stateful JPD =
instances waiting on say different events that would then make them "runnin=
g", what would be the impact performance wise on the container? (this is no=
t a naive question...I am well aware that containers are optimized to handl=
e high loads but this is a JPD instance and therefore its representation as=
an EJB is somewhat opaque).
I understand that a stateful JPD is an entity bean...and given the fact tha=
t I avoided them like the plague, it is difficult for me to say what the pe=
rfromance impact would be, especially CPU-wise...memory is not a concern bu=
t CPU usage of thousands of EJB entity beans would be of great concern.
Any helpful hints (apart from prototyping and/or stress testing) would be v=
ery much appreciated.
regs
beez
-
Re: Internals of a stateful JPD and impact of having 1000's of them
If CPU usage is a concern, i would think that not the number of instances w=
aiting but the frequency of the events (that make them running) would deter=
mine how your application perform. I could have had created 1000 instances =
of a process with them waiting for different events. Once they are created =
and are waiting, CPU usage is not significant. Now as and when the events s=
tart coming in, the container loads the entity beans and invoke operations =
on them. So if I have a figure like 100 events/min coming in, the CPU usage=
is quite higher than if I have 5 Events/min coming in.=20
Also, note that i may not be only the entity bean using your CPU. It might =
also be the AsyncDispatcher and the JMS service that dispatch the events wh=
o would want their share of CPU to dispatch the events to the waiting proce=
sses.
-
Re: Internals of a stateful JPD and impact of having 1000's of them
Hi Rajav
Thanks for a very succinct and logical reply.
It makes absolute sense that it is the events that
the instances are waiting on, that would determine the performance of said =
waiting processes.
regards
k