> > Do you think we need more NIO work on trailers etc? It > would seem that > > nodes accept a bunch of trailers and then overload for a while on > > threads - so effectively we DO have a bandwidth based QR, > it's just a > > delayed effect. This may be bad for routing... > > I find it hard to believe that the problem could be due to a lack of > NIO, given that this problem did not occur back when we had no NIO > whatsoever. This isn't to say that NIO doesn't help, I just > feel that > it is unlikely to be the crux of the problem.
A couple of thoughts on this.. Routing doesn't work very well -> request chains becomes longer than they where when routing worked... Also, (possibly an effect of the above).. The probability of a sendDataFailed seems to be higher than the probability of a receiveDataFailed. Combined with the fact that nodes seem to continue to download data even though the requestor they are relaying to has died this might well cause larger amounts of trailertransfers going on in the network. The two issues above both causes more threads to be involved for any given request that manages to locate some data. So.. Back to square one again.. Routing problems.. /N _______________________________________________ Devl mailing list [EMAIL PROTECTED] http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/devl
