Anyway I think most tutorials only work in windows especially the thing that a key mostly is writeable in linux will make you to hack some workarounds to not have zillions of wakeups on write events.

I will try this tutorial. I’m writing since 4 years (yet another) caching framework (which I will release on SF) and will include a nonblocking distributed server/client cache. I have a total well performing blocking solution with worker threads, traffic control etc but I want to measure the difference between a well implemented nonblocking solution compared to a blocking.

It will not be easy since serialization is involved and I think I will be needing a bunch of worker threads handling send queues. People tend to say that you always should use one thread but I’m skeptic since
the underlying operations might read a file, db or so=block.

Everywhere I read people says Serialization=not NIO. but I know it can be done since I have a stupid echo serializing server now.

Keep up the blogging about this issue I sure need it

Kindly

//Marcus

I found that a pretty nice package has been moved into Apache called MINA. I’ve decided that I will most likely use this for our purposes. It was previously called Netty2. If you don’t feel like struggling with a poorly documented API unnecessarily, I’d consider it. MINA is still lacking documentation, too, but the mailing list is very lively, and the examples are pretty good.

I’ve also read that under certain conditions/requirements, using BIO still wins out in performance. I’ll be darned if I remember where, but I think it was on Sun discussion boards, or maybe the IBM ones.