[ale] Co-lo woes

djinn at djinnspace.com djinn at djinnspace.com
Sat Oct 6 10:59:02 EDT 2001

Well, happy Saturday morning to you all.  I need some assistance from
any network analysis gurus out there to help me determine what is going
on with my co-located network.

Some history:  I have been observing 35-45ms ping times and avg. 30%
packet loss on my machines located at IBM.  Since we came off a network
with 1000ms ping times and 50-60% loss, I never complained (I also
suspected it was my firewalling system, but had no real way to test that
theory).  IBM had some catastropic VPN failure last night, and lo, there
is now 0 - 3% packet loss and and avg of 19ms ping times.

But my web sites are still exhibiting the same behviour they were during
the IBM outage -- I can contact the site, DNS is fine, but anything that
must transmit over multiple packets seems to just die.  Large images,
large blocks of text, anything like that.  Now, it wasn't doing this
yesterday morning before IBM started and the only thing I've done
differently is reboot the firewalls, which usually helps my throughput
(or seems to, but again, I don't know how much of my historic problems
have been me or IBM).

The other thing the web sites are doing is they'll start out around
15kbps transfer...and then go to 10...and then 9...8...7...it's like
watching a shuttle countdown.  And then they get to 1kbps, and they

IBM is trying to blame this on me.  It may in fact be me, and that's
fine.  But if it is me, I need to fix it.  And if it's not me, I need to
prove it.  I'm running iptraf on both firewalls and everything looks
more or less normal.  I'm also running ntop on my external router.

If anyone wants to observe this behaviour for themselves, the pirmary
site in question is http://www.scubadiving.com.  All our sites do it,
but scuba is the biggest and ugliest and worst offender for massive
images...and it's *never* behaved this way before.

So.  What suggestions are out there for what I can do?  I'm going to try
to measure the load time by loading lynx on the same subnet and
comparing it to lynx from here.  Are there any other things I can do to
help me determine where the problem is??


This message as been sent through the ALE general discussion list.
See http://www.ale.org/mailing-lists.shtml for more info. Problems should be 
sent to listmaster at ale dot org.

More information about the Ale mailing list