openbsd-misc
[Top] [All Lists]

Re: Optimizing Bridge Performance

To: Theo de Raadt <deraadt@cvs.openbsd.org>, misc@openbsd.org
Subject: Re: Optimizing Bridge Performance
From: Mike <owensmk@earthlink.net>
Date: Wed, 29 Mar 2000 07:13:49 -0600
References: <200003291835.LAA12474@cvs.openbsd.org>
Sender: owner-misc@openbsd.org
Sorry, am using 2 Intel EtherExpress Pros. But let me preface this with
an apology---my network perfomance is much better than I originally
thought.

I sent a post to Luke, meaning to post it to the list that said that I
was wrong about the bridge as the culprit---the NAT from within the LAN
worked fine, which means that it must have been something to do with the
SSH scenario when I logged in from outside. Right about the time I sent
that message, I tried SSH again (from a machine outside the network)
---and everything was fine, great, perfect. So then I posted the "let me
backpeddle harder" message to you, meaning again to post to the list.

Now here is what I think happend, and I apologize for the abysmal
complexity of this whole ado: This all began last night --- the bridge
setup. After I set up the bridge, I went home, knowing that it worked ok
(b/c I could ping out).

I live in Fort Worth Texas, and you may or may not know that we had a
tornado last night---5 minutes from my apartment. It busted downtown to
pieces. Well after the storm was over, happy to be alive, I couldn't
resist testing the bridge. So I dialed up and SSHed in. This was the
first real delay---major slow, in fact nothing happed after the
password---dead stand still. When I got to work today, thinking it might
have been an oversight in my firewall rules. Setting the bridge to 'pass
in all', I tried it again---same thing: dial in from my laptop and SSH
in. Same thing, but at least I got passed the password. Still very slow.
I read the man pages in greater detail, trying to do my civic duty, and
after no luck, I posted a message.

Well it never occurred to me that it was probably the ISP's loop that
was dragging. I am almost sure of it now. The ISP is in downtown Fort
Worth.

Over the last 20 minutes, I have gotten normal performance---with NAT
and with SSH. The very fact that I got this performance suggests that it
has nothing to do with the bridge, ethernet cards, SSH, or other.

It was the tornado. Of all variables. Thus, the policy from now on when
confronted with a problem: read the man pages again, check the whether,
then post.

Thanks.



Theo de Raadt wrote:
> 
> what ethernet cards are you using?
> 
> cards with small fifos or long interrupt handling sequences suck for the 
> bridge
> code, because of how it is designed.

<Prev in Thread] Current Thread [Next in Thread>