You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mina.apache.org by Jeff MAURY <je...@jeffmaury.com> on 2013/10/15 16:51:37 UTC

Blog article on Netty 4

Hi,

an article has been published about the design of Netty 4. I think some
topic maybe of interest.
https://blog.twitter.com/2013/netty-4-at-twitter-reduced-gc-overhead

Regards
Jeff

-- 
Jeff MAURY


"Legacy code" often differs from its suggested alternative by actually
working and scaling.
 - Bjarne Stroustrup

http://www.jeffmaury.com
http://riadiscuss.jeffmaury.com
http://www.twitter.com/jeffmaury

Re: Blog article on Netty 4

Posted by Emmanuel Lécharny <el...@gmail.com>.
Le 10/15/13 4:51 PM, Jeff MAURY a écrit :
> Hi,
>
> an article has been published about the design of Netty 4. I think some
> topic maybe of interest.
> https://blog.twitter.com/2013/netty-4-at-twitter-reduced-gc-overhead

That's interesting, but I would rather avoid allocating buffers myself.
The real problem is that teh GC is killing the performances, and even if
you manage your own pool of buffers, you still have GC pauses. GC pauses
are unavoidable, the only thing you can do is to try getting less
pauses, by using G1, for instance.

Now, I think tehre is a better solution : using a pre-allocated buffer
associated with the thread (ThreadLocalStorage). If we need to copy the
containt of this buffer, fine, but I'm quite sure that most of the time,
it will not be necessary.



-- 
Regards,
Cordialement,
Emmanuel Lécharny
www.iktek.com