Hi Brian, OS: Red Hat Enterprise Linux AS release 4 (Nahant Update 2) Kernel 2.6.9-22.ELsmp on an i686 Xalan version: 2.7.0 Xerces Version: 2.9.0 JDK: 1.5_11-b03 server edition Min-Max memory: 3 GB
Please, can you give me an insight on how you and Toadie tracked down the performance related issues? Also, I am not able to understand the performance differences in XSLT processing between OS. With linux box having superior hardware Vs Windows desktop, we would expect a better performance in linux, but surprisingly it is the reverse! Meanwhile, I will also try with Xalan 2.7.1 and capture the difference it makes. Regards, Sriram Brian Minchau wrote: > > > Sriram. > I can comment on that link (my name is in there somewhere!). > > Yes, the SUN JRE had some horrible synchronization performance "bug". It > made it really excellent for finding code in Xalan-J that did > synchronization, and cleaning up many of the places that used synchronized > methods, but didn't have to. I worked with Toadie and flushed out many of > those spots. Those changes all went into Xalan-J 2.7.1 > > Which JREs are you using, and which version of Xalan-J? > > Still, the pressure needs to go to SUN to fix how it handles > synchronization. > > - Brian > > > > > Sriram83 > <[EMAIL PROTECTED] > com> To > [email protected] > 09/16/2008 10:47 cc > AM > Subject > Re: XSL Transformation in Xalan - > Pls help! > > > > > > > > > > > > Michael: yesterday, we stripped the xslt transformation component > seperately > and executed the xslt transformation for about 1000 transactions. But > still, > there seems to be difference between execution time in windows vs > execution > time in linux box. > > Also, we came across this post regarding Xalan in multicore environment.. > Can anyone provide comments on this post? > > http://marc.info/?l=xalan-j-users&m=117327580321384&w=2 > > Regards, > Sriram. > > > Michael Ludwig-4 wrote: >> >> Sriram83 schrieb: >>> 1/. We have the XSL transformation taking place inside an application >>> deployed in JBOSS app server. There are no restart of JVM. >>> Initialisation of memory takes only during jboss startup. >> >> I don't know much about JBoss, but it's a pretty huge application. You >> may be testing JBoss instead of Xalan. In order to test Xalan, I'd >> rather use a leaner frame, maybe just the simplest of programs which >> does what you describe below using loops and timestamps. >> >>> 2/. As mentioned by Tatu, I believe we do allow time for JVM to >>> warmup. We run around 200 XSL transformations first, after this, we >>> collect the average time taken for around 1000 transactions. >> >> Sounds very reasonable to me. >> >> Michael >> >> > > -- > View this message in context: > http://www.nabble.com/XSL-Transformation-in-Xalan---Pls-help%21-tp19471054p19513437.html > > Sent from the Xalan - J - Users mailing list archive at Nabble.com. > > > > > > -- View this message in context: http://www.nabble.com/XSL-Transformation-in-Xalan---Pls-help%21-tp19471054p19515866.html Sent from the Xalan - J - Users mailing list archive at Nabble.com.
