I didn't expect that kind of difference either, but I ran the tests several times yesterday with the same results. I'm running the same tests again today, but our LAN is on the fritz so I'm getting inconsistent results. oh well.
thanks for the explanation on the relevant improvements in jdk1.4 which contribute to the performance enhancement. peter Remy Maucherat wrote: > > peter lin wrote: > > Some interesting performance results. I found some time to test my > > webapp a bit more with 4.1.7 and discovered a performance difference. > > > > when I benchmark my webapp using 4.1.7 w/ jdk131 on solaris 8 I get an > > average response time of 490ms, and a low number of 460ms. when I swith > > to jdk141, I see a 10-20% performance boost. the only thing different > > between the two tests is the change in jdk. It's pretty dramatic > > considering the fastest response time using jdk141 is 380ms. > > I can confirm that JDK 1.4 provides better performance with Tomcat 4.1. > > JDK 1.4 improves quite dramatically b2c and c2b conversion (it uses new > implementations from the nio package transparently). It may also improve > networking performance. > > Those two items are the biggest hotspots of Coyote (and can't be > improved much), so that's why you get a performance boost. > > BTW, it's a pretty impressive gain for a real world application. I > thought you would only see that kind of gain with synthetic benchmarks. > > Remy > > -- > To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> > For additional commands, e-mail: <mailto:[EMAIL PROTECTED]> -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
