Hi,
1. has somebody maybe a little graphical sketch about the logical
structure and the responsibilities of the single components within the
Mina queue? I know there is one document, which describes the data flow,
but I'm missing the key, which parts are (maybe) concurrent and which
parts are serial (one thread). What I'm searching for (because I'm a
little bit confused about it at the moment) is the following:
- acceptors: Do they split into more than one thread? I think so, right?
Where I can adjust the parameters for this behaviour?
- filterchain / codecs: Is there more than one concurrent chain? Depends
from acceptor threads, right? Is it right, that I can't use lesser
chains than acceptor threads? Can I add more chains by adding a
ThreadPoolFilter at the beginning?
- sessionhandler: How many sessionhandlers exist per default? One or as
many as filterchains? Can I modify this?
It would be really great(!) not only for newbies to have such an sketch
to go sometimes back and remember, how all works together.
2. I've implemented my current application in the normal way, that the
codecs decode the data and forward them to the sessionhandler, where
they are handled. Now I'm wondering, if it would also be possible (and
clean from the view of Mina) to handle the decoded data already in the
codecs itself without forwarding them to the sessionhandler. Would I get
a performance impact in this case?
I hope that somebody has a little bit time to answer my questions.
Best Regards
Michael
