Subject: slip problems under high load
To: None <current-users@NetBSD.ORG>
From: Robert Dobbs <email@example.com>
Date: 11/22/1994 19:50:54
this on an i386 running a 1.0 kernel with cslip (link0 and link2)
when I access news (highly compressible text) across my link, I
get ~6000 cps and concurrent ping time of between 400 and 900ms
this seems fine
when I access large items over WWW (mosaic, netscape, or lynx)
depending on the data I get between 3000 and 6000 cps, but ping
times just to 1000ms initially and climb upwards until they
are ~17000ms and higher.
This is a connection to a sgi indigo on a local ethernet on the
far side of our slip line. As this line clog doesn't seem to occur
with hosts further away.
My conclusion is that something about the tcp on the WWW machine
I'm connecting to is broken such that it doesn't listen to window size
requests. Is there anything netbsd should be doing in cases like this?
Can you do anything about data flood?