On Wed, 14 Nov 2007 21:22:48 -0500
Sam Hartman <hartmans-ietf(_at_)mit(_dot_)edu> wrote:
Joe> By essentially shutting your machine down for over an hour.
No, not at all; not even close.
I'm only going to send this one message, but then I'll drop out of the
thread. We've drifted far from Leslie's original query.
Steve did not say his machine was CPU bound. Also, even if it was CPU
bound, he's probably running an operating system with reasonable
multiprogramming characteristics. So, if he wanted to use the
machine, his backup would take longer, but he'd get to do whatever he
wanted.
Yes, Steve almost certanily did slow down any heavy CPU use during
the time when he was doing the backup.
Right. Operating systems are really good at sharing the CPU between
processes. As long as you're talking about pure CPU load, the
scheduler works *very* well, and has for O(45 years). Right now, as
I'm typing this, I'm running that dump again in the background, this
type piped to 'cat >/dev/null' to better simulate the pipe overhead,
and I'm running 'openssl speed' in another window. I do not perceive
any slowdown whatsoever. I would notice it I tried to view a movie, or
if I tried to do serious picture or video editing, or if I tried
running another command -- say, firing up firefox if it were not
already running -- that did a lot of disk I/O. I don't know of any OS
-- and that includes every version of Unix I've ever used, going back
more than 30 years, and every version of Windows I've used, going back
about a dozen years, that shares disk bandwidth particularly well. But
sharing CPU is very easy.
I don't dispute Joe's numbers about how crypto can limit bandwidth.
Again, though, what matters to me is system and application
performance, and there I generally don't feel the hit.
Why do I use GigE? Well, in one sense it's because that's what came
with my computer, but I did spend the money to convert my house network
to GigE. There's some unencrypted traffic to the file server. Just as
important, even with the crypto I'm approaching the limits of what I
can do with 100BaseT. As I noted, my actual, over the wire, dump
operations are running at 75-80M bps. There's not a lot of headroom
between that and the 91M bps that is the fastest I've managed with
ttcp. I just got a new, faster laptop; I'm curious what it will do,
with and without the crypto.
--Steve Bellovin, http://www.cs.columbia.edu/~smb
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf