Trustin Lee wrote:
Why don't we use pooling techinique? We could use commons-pooling or
our proprietary stack implementation. :)
What you talkin bout Willis? :) Proprietary?
2005/11/19, Emmanuel Lecharny < [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>>:
> >Added a brutal and rude synchronization to avoid a race
condition in DnParser usage.
> >
> >At least, it corrects the problem, but the server is now 3
times slower.
> >
> >
> Darn I was enjoying the recent speed ups too. We need to figure
this
> one out.
Well, after having tuned the log level, the penalty is not that
awfull :
something like 30%.
I got 113 search requests per second, against 143 req/s four month
ago.
There are huge areas of optimization to work on, but from now one,
I bet
the first move is to finalize the 1.0 beta. We don't need a
blazing fast
server which is full of bugs ;)
+1. I think we should add features up until we get to a beta. Then we
can spend our time optimizing. Who knows what will happen to your
optimizations as new features are stacked on. We might break them I
mean and you will have spent good time on it. Then again I am enjoying
the performance boosts we have gotten from your recent work. Thanks!
Alex
