Yes, if the iohandle is blocking the execution then that would obviously do it That said I am not sure how the reactor threading in Mina would handle very small messages sent in a way that prevents the data from aggregating. Difference is 20k individual triggers or 1k triggers for the same number of messages. On Aug 15, 2014 2:00 AM, "Emmanuel Lécharny" <[email protected]> wrote:
> Le 15/08/14 05:41, Jon V. a écrit : > > Emmanuel, is there a line test tool for Mina? I ask because seems like > the > > Jedi Jin here has a throughput problem. > > Something like > > http://blog.softlayer.com/2011/using-iperf-to-troubleshoot-speedthroughput-issues > should do the trick. But here, with the client and the server o the same > machine, I doubt the network could be the probelm. > > > > Jin, > > Have you tried benchmarking your code directly? In all reality I process > > 30k sockets and 500mb/s using raw TCP in java so if there is a problem it > > is either your code or Mina. > > It's crystal clear that it's not a MINA problem per se. I'm able to > saturate a 10Mb/s network with more than 20 000 messages per second on > my computer on my LDAP server using MINA. > > One of the possible reasons : the server IoHandler is doing weird things > (like waiting for an external resource before releasing the thread). > Typical use case : sendinhg the received data to a remote data base. > > So bottom line : what do you do on the server side with the data you > receive ? > >
