Trustin Lee wrote:
Are you saying that it would be great if you can switch the mode (semi-synchronous stream I/O and original MINA way)? If so, this is a wonderful idea.
That is exactly what I'd *like* to do, but I lack sufficient depth in asynchronous architecture to know whether it's feasible. I was hoping for some opinions. :)
BTW, it's conventional wisdom that apart from handling keepalives efficiently, asynchronous I/O isn't strongly beneficial for Java servlet systems, since each thread is either 100% busy doing "servlet business" or 100% idle in the pool.
However, I think there are several "waity" areas that would benefit from async. You shouldn't have to block a thread waiting for a slow backend in a proxy-like system, or when you're waiting for an authentication response from a heavily loaded LDAP server. It probably makes sense for slow clients. It might even make sense to implement semi-sync database drivers someday! :)
In the long run, I'd like to employ a heuristic approach that would switch, based on observed latency, back and forth between blocking and non-blocking I/O on each connection (e.g. go async when latency is > 100ms?). I like how MINA's modular architecture makes it fairly easy to experiment with different approaches. I hope it won't add too much latency. :)
-0xe1a -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
