Hi, As of Mustang build 59 the NIO Selector implementation uses epoll when running on 2.6 kernels: http://blogs.sun.com/roller/page/alanb?entry=epoll
Would be nice to see some figures for MINA using this build. BTW, we should have a benchmark test suite for MINA. That way we can easily see improvements after optimizations or compare different platforms/technologies. There's already a JIRA issue for it: https://issues.apache.org/jira/browse/DIRMINA-56. Since you guys already have some experience with benchmarking MINA and NIO it would be great if you could add some thoughts (as comments in JIRA) on what such a test suite should include. /Niklas [EMAIL PROTECTED] wrote: > Hi, > > Yes, I am pretty sure that the JDK does not currently have epoll support - > there is no Sun JDK for the 2.6 kernel specifically. > > However, I understand that JDK 1.7 is going to introduce AIO support - > which doesn't help us right now unfortunately! Given that Solaris has very > good AIO support I can't understand why Sun haven't done this sooner. > > In our testing on RHEL 3, on a 4 way dual-core opteron box, we have found > that the BEA JRockit performance (measured in throughput) is about 5-10% > faster than the Sun JVM, including the latest JDK 1.6 beta. I am guessing > that their NIO implementation is a bit better than the one in the Sun JVM > since our test doesn't really exercise much else. > > Robert > > > > > "Vinod Panicker" > > <[EMAIL PROTECTED] To: > [email protected] > m> cc: > > Subject: Re: Performance in > MINA > 27/02/2006 18:11 > > Please respond to > > mina-dev > > > > > > > > > > Hi Robert, > > On 2/27/06, [EMAIL PROTECTED] <[EMAIL PROTECTED]> > wrote: > >>Hi, >> >>This is maybe a suitable thread for me to discuss some of the performance >>analysis that I've just started doing. > > > <snip/> > >>I have also implemented an "AIOSocketIoProcessor" using the IBM > > alphaworks > >>aio package (which is not open source but useful nonetheless to determine >>whether it is worth building an AIO library). Even using the naive >>threadpool that comes by default with the IBM library, latency was > > improved > >>significantly on win32. I was going to do further testing with the IBM > > AIO > >>library but unfortunately it doesn't support the linux version we run on >>our larger boxes (RHEL 3). Does anyone know of any other AIO libraries > > for > >>Java? > > > Seems like the IBM aio4j is the only true aio library out there. > Really unfortunate that they dont have an open source license for that > inspite of many repeated requests. My personal findings with java nio > are that there is significant room for improvement. The current nio > implementation is a cpu hog, even when it gets to a huge number (over > 10K) of idle connections. I dont think they have implemented epoll > support as yet in the jdk - something that blackdown had done with > theirs. > > Also the aio4j library doesn't support SSL/TLS :( > > An open source aio implementation is something that is begging to be > done, and is on my long term agenda. > > >>Minimising Copying: >> >>One of the operations our servers do is take in data and retransmit it >>unmodified to a number of subscribers as quickly as possible. Therefore >>minimising copying of data is important for us. > > > zero data copy is the holy grail for network servers. It would be > great if we can minimize all the data copies that are happening. > Would increase performance by many a magnitude. Would be eagerly > waiting for your proposed changes. > > <snip/> > >>We are going to be doing performance testing with larger numbers of >>connections (e.g. 500) later this month and I will report on my findings. > > > I'm looking at testing for much higher numbers (100K+), but > unfortunately its always been put on the back burner. Someday! > > Regards, > Vinod. > > > > > > This communication is for informational purposes only. It is not intended > as an offer or solicitation for the purchase or sale of any financial > instrument or as an official confirmation of any transaction. All market > prices, > data and other information are not warranted as to completeness or accuracy > and > are subject to change without notice. Any comments or statements made herein > do not necessarily reflect those of JPMorgan Chase & Co., its subsidiaries > and affiliates. > -- Med vänlig hälsning Niklas Therning Software Architect [EMAIL PROTECTED] Mobil: +46 739 75 05 73 Trillian - Software Design at its best www.trillian.se
