Hi, I spent some time investigating mina performance. I've written two simple multiplexing servers that collect 16 bytes from one connection and send it to all other. One was using mina and another was using NIO directly. Comparison program collects avarage lag in message delivery depending on number of connected users. Dummy server was performing few times faster then MINA. It was a real surprise for me. There are comparison chart and sources of my tests attached to this letter. Chart data can be reproduced using command: java -jar TestPrototype.jar <server-ip-or-dns> 55555 20000 1000 1 200 10 MINA is really usefull library, but if my tests is ok it can become performance bottelneck. Any way MINA can and should perform faster.
Tim