geronimo-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Zakharov, Vasily M" <vasily.m.zakha...@intel.com>
Subject SPECjAppServer2004 details (was: Some observations..!!)
Date Mon, 22 May 2006 16:47:20 GMT
Hi, Santosh,

> JOPS ~ 1 (for) Geronimo looks very less from ur report.

Sure, it's due to some configuration problem or bug.

SjAS logs clearly indicate some benchmark components are not working as
expected.

> Does SPecjAppServer2004 might include some ejb (entity) transactions?

Sure.

> Do u have any numbers in a itemized way (more refined than JOPS
metric).

Sure.

> Pls post them, that would be better [ atleast for me :-) ]

Here I attach the whole output directory contents.
You may compare them with the details published in FDAs here:
http://www.spec.org/jAppServer2004/results/jAppServer2004.html

> I am also testing Geronimo for our sample app.

Wow, it's great! Performance is important indeed.

Aren't you planning to use SjAS for measurement?

 Vasily


-----Original Message-----
From: Santosh Koti [mailto:Santosh_Koti@infosys.com] 
Sent: Monday, May 22, 2006 6:46 PM
To: user@geronimo.apache.org
Subject: Some observations..!!



Vasily,

Great effort!!


  JOPS ~ 1 (for) Geronimo looks very less from ur report.

  Does SPecjAppServer2004 might include some ejb (entity) transactions?


  Do u have any numbers in a itemized way (more refined than JOPS
metric).


  Pls post them, that would be better [ atleast for me :-) ]


  PS:   ( I have not done SpecjAppServer benchmarking )

 


I am also testing Geronimo for our sample app. So far the results are
quite impressive for pure data fetches , but those involving
transactions are not very appreciable. (I am not sure; is it to do with
any config, though I have tuned all my infrastructure components)

   Roughly speaking (in short):
           My readings (After some bit of tuning) for my sample app:
                           My use case was combination of some txns &
some  				   pure data fetches.

                       Run Duration: half-an -hour (will increase later)

                       Users: 225

                            1. Pages Per Sec (Avg): 182.518

                            2. Response Time  (Avg): 0.123 sec
                            3. App Server-CPU% utilisation: 15.66

                            4. Web Server-CPU%: 5.461

                            5. DB Server -CPU%: 10.088                 


            At a load of 250 it was breaking down , I increased the heap
size to 1.5 GB, but the DB process exceeded beyond the (500-configurd ,
by DBA) mark.



EJB transactions commit take too much time, & in doing so keeps the db
threads alive for undesired amount of time, leading to an undesired
response time

 ^.-(



Thanks,
Santosh.
"Don't talk about yourself; it will be done when you leave. "



-----Original Message-----
From: Zakharov, Vasily M [mailto:vasily.m.zakharov@intel.com]

Sent: Monday, May 22, 2006 6:45 PM
To: user@geronimo.apache.org
Subject: RE: SPECjAppServer2004 up!

Matt,

Sorry for the delay, I missed your question.

Sure, all the DDs and the deployment process description are available
at:
http://opensource.atlassian.com/confluence/oss/display/GERONIMO/SPECjApp
Server2004

In fact, not everything is smooth now. The SPECjAppServer2004 Driver
starts and completes normally, however the run logs show that only
Orders Transactions work normally, while Work Order, PO and POLine
Transactions don't in fact work. I guess some configuration issues still
exist, while not clearly visible. I can also post the output directory
logs or the whole FDA if it's useful.

 Vasily


-----Original Message-----
From: Matt Hogstrom [mailto:matt@hogstrom.org]

Sent: Saturday, May 13, 2006 1:20 AM
To: user@geronimo.apache.org
Subject: Re: SPECjAppServer2004 up!

Awesome news :)

Can you share the DDs ?

Zakharov, Vasily M wrote:
> Hi, all,
>
> Finally, after a long struggle, SPECjAppServer2004 benchmark is up on
> Geronimo 1.0!
> (on my notebook :)
>
> My thanks to Matt Hogstrom, David Jencks, Jacek Laskowski and Aaron
> Mulder for their help with various problems and issues that occured in
> this work.
>
> Now I'm working on updating the running manual and on creating a
> reasonable hardware configuration to measure the actual performance
> more effectively.
>
> Vasily Zakharov
> Intel Middleware Products Division
>


Mime
View raw message