On Tue, 2013-05-07 at 11:15 +0100, ant elder wrote: > On Thu, May 2, 2013 at 1:02 PM, Oleg Kalnichevski <[email protected]> wrote: > > Folks, > > > > I invested some time into performance optimization of non-blocking core > > components in order to push HttpCore 4.3 and HttpAsyncClient 4.0 closer > > to a GA release. I implemented several local optimizations in the low > > level transport code and also fixed a bug in the non-blocking connection > > pool code that can be a major performance inhibitor when executing > > thousands of small requests in a tight loop. > > > > My performance benchmarks, admittedly very limited and > > non-comprehensive, now show pretty good results in client side test > > scenarios for NIO based implementations. In fact, in my tests HttpCore > > NIO performance matches that of blocking HttpClient even with a very > > modest number of concurrent connections (<20) which is something I > > thought I would never see. > > > > Thats great, i'm really impressed you've been able to get the NIO > performance so good. >
I only fixed a couple of bottlenecks but otherwise most of performance gains are probably due to improved epoll selector implementation in the latest JREs. Oleg --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
