freebsd router

[Danial Thom]

--- "Matthew D. Fuller"
<fullermd at over-yonder.net> wrote:

> On Wed, Jan 11, 2006 at 06:10:15AM -0800 I
> heard the voice of
> Danial Thom, and lo! it spake thus:
> > 
> > Well thats just stupid. Clearly I'm referring
> to general purpose
> > operating systems, not custom hardware
> platforms.
> 
> "[...] fastest router platform man has ever
> created" doesn't include,
> imply, or even allow such qualifications. 
> State insane hyperbole,
> expect to get called on insane hyperbole   8-}
> 
> 
> > And I don't see any of your logic for
> thinking that "newer FreeBSD
> > versions" are faster. They're not.
> 
> Well, for starters, _I_ never expressed any
> opinion as to whether they
> were.  5.x certainly isn't (but then, 5.x is
> slower at most things
> than 4.x).  6.x may be, in certain situations. 
> And 7.x very likely
> will be, particularly with the work Andre is
> currently doing in the
> network stack.  MP is a DECIDED advantage in
> routing since you can
> process things in parallel, and routing is a
> _very_ parallelizable
> operation, since packets are independent of
> each other.

No, that's wrong. Firstly, you CAN do things in
parallel, but when you chop up the "tasks" in
routing you don't gain anything, in fact you
lose, because it is best done as a single task.
Think of it as a 100M dash with 1 guy against a
relay team of 2 guys, with all 3 exactly the same
speed. The 1 guy will win every time, because
you'll never overcome the hand-off. The best
you'll ever be able to do is tie him. By
"threading" the kernel you make it run more
efficiently when you have tasks that have to
stop-and-wait for other things, but in a router
your main, preemptive task is the routing, which
you can't improve by parallelizing it. In fact
you add lots of unnecessary cpu cycles that get
in the way of the single task approach. You
*might* be able to gain additional capacity IF
you design the rx/tx tasks in a certain way, but
you'll have more latency through the box.

I can promise you that 7.0 is far from "probably"
anything. They are a long way off. Maybe come up
with some real benchmarks (like the one I
suggested in my other post) so you won't be
bamboozled by the hype so easily.

DT

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com