Well in my initial tests, I did a smaller query first to warm things up.
But it was really only a quick run of each, as I had other things to do.
Unfortunately, this real-world tests may have led me to an erroroneous
conclusion. My further investigations today still did not have the
controls and repetition for definitive results, but they are at least a
little better.
What I did for each was start JBoss, run the test twice, log off and
wait 5 minutes, test twice... repeat...
I did have a couple of poor runs at the beginning, but I now believe
this to be an OS issue(memory paging?), rather than a VM issue (it
occured this time on the client VM).
Here's some of my runs:
(results in seconds)
Client VM Server VM
(Load)
Run1 25 22
Run2 17 15
(Pause)
Run3 43 18
Run4 25 13
(Pause)
Run5 18 13
Run6 19 18
Median 20 17
Average 24.5 16.5
The server definately takes longer to display the first JSP page (not
timed), but it doesn't show any disadvantage in the first search. Seems
my original conclusions were misleading, and the server VM may be a bit
faster.
> Also have you tried running it on the Jrockit JVM? It consistently
> outscores sunVMs on appserver profiles.
BEA JRockit? Hmm that would be interesting to compare, but I highly
doubt I could convince my boss to use a JVM other than Sun's! ;)
As it stands, performance seems much more acceptable than that
off-the-cuff test indicated.
Thanks for your responses,
Jeff Bischoff
Kenneth L Kurz & Assoc, Inc.
Dhananjay Prasanna wrote:
------------------------------------------------------------------------
*From:* Jesse Sightler [mailto:[EMAIL PROTECTED]
*Sent:* Saturday, 24 June 2006 11:36 AM
*To:* MyFaces Discussion
*Subject:* Re: [OT] JSF and JVM performance
How many times did you run your performance test? Were you timing the
first run, or timing some runs after giving it some stress and then a
quick breather? :)
<snip>
On 6/23/06, *Jeff Bischoff* <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
I have noticed that Myfaces seems to perform better while running on the
Hotspot Client (default) jvm, rather than the Hotspot Server jvm. Have
any of you noticed this too?
I would be very interested in this test too, as Jess says more detailed
metrics would really help.
Also have you tried running it on the Jrockit JVM? It consistently
outscores sunVMs on appserver profiles.
Jeff Bischoff
Kenneth L Kurz & Assoc
This correspondence is for the named persons only.
It may contain confidential or privileged information or both.
No confidentiality or privilege is waived or lost by any mis transmission.
If you receive this correspondence in error please delete it from your
system immediately and notify the sender.
You must not disclose, copy or relay on any part of this correspondence,
if you are not the intended recipient.
Any opinions expressed in this message are those of the individual
sender except where the sender expressly,
and with the authority, states them to be the opinions of the Department
of Emergency Services, Queensland.
