[openbeos] Re: Networking : R5 or BONE
- From: "Nathan Whitehorn" <nathan.whitehorn@xxxxxxxxxxx>
- To: openbeos@xxxxxxxxxxxxx
- Date: Sat, 26 Jan 2002 09:52:38 EST (-0500)
> > Since BONE is obviously vastly superior, and largely (completely?)
> > backwards
> > compatible, it seems silly to implement the (rather lacking) r5
> > stack
> > when
> > there is the foundation for something much better.
>
> That networking isn't as good as it could in R5 doesn't have anything
> to do with the fact that there is a net_server. If the net_server
> would
> be reimplemented, there would be no need to include all its bugs with
> it.
> Anyway, the network team is preparing a kernel-level stack, similar
> to
> BONE, to avoid the context-switch overhead (which can be huge for
> networking stacks [lots of very small packets]).
Really? It says in the FAQs that they are staying in user land.
Besides, much of the overhead with net_server is not context switch
overhead. app_server has the same kind of ops and low latency. What
killed net_server's performance was the truly enormous number of copies
(I think it was in the double digits) that it made of *every* piece of
data that went through it. BONE was a one copy (memory->net card)
stack, and was made as close to zero-copy as humanly possible, which is
responsible for much of its performance.
-Nathan
--
Fortune Cookie Says:
Bumper sticker:
"All the parts falling off this car are of the very finest British
manufacture"
Other related posts:
