nel-all
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Nel] NeL Network Engine


From: Leighton Haynes
Subject: Re: [Nel] NeL Network Engine
Date: Thu, 1 Mar 2001 10:45:11 +0800
User-agent: Mutt/1.2.4i

On Wed, Feb 28, 2001 at 11:13:12PM +0100, Bernard Hugueney wrote:
> * Zane <address@hidden> [010228 19:17]:
> > ----- Original Message -----
> > From: "Vincent Caron" <address@hidden>
> > Subject: Re: [Nel] NeL Network Engine
> >
> >
> > Would not the wisest choice be to research all the high performance web
> > servers out there?  It seems to me that this problem has been researched
> > quite extensively by a LOT of other GPL projects.  I doubt you'd be able to
> > find a better solution than what Apache or other heavy duty web servers have
> > already implemented.
> >
> > -E.J. Wilburn
> > address@hidden
> >
> Indeed, this is a MUCH debated suject. But I'm afraid Apache is not
> the fastest example to follow (they have special constraints such as modules
> interface) Zeus and TUX Threaded linUX webserver or BOA come to mind.
> http://www.kegel.com/c10k.html seems interesting but a nit out of date ...
> Anyway, there was a discussion on lkml on how to overcome select()/poll()
> shortcomings with a new API. If it made it to kernel 2.4, you should be
> looking that way.
> 
<troll>
Hey! Why don't we shove the networking engine in kernel-space! Then we
can stop all these other pesky processes from pre-empting us and stuff.
</troll>
I'm not sure if this is a really important issue right now. As many
people have noted, premature optimization is the source of all evil. 
What we _need_ is a good api to the networking code, so that the 
underlying code can be optimized later. Actually optimizing the way it works
should be dealt with when you need to deal with it. After all, it may work out
that the control code behind it all ends up needing 1/4 of a cpu/thread, at 
which point this problem morphs into a completely different one. 
Someone might release a nice lightweight threads architecture in 4 months
meaning the point is unimportant. 

Leighton...

--

Part-time student. Full-time Programmer. 
Seeking the 36 hour day and the 10 hour working week.
(08) 9272 9058 (Home - like I'm ever there)
0401 335 136 (Mobile - like it's ever on)


reply via email to

[Prev in Thread] Current Thread [Next in Thread]