We've noticed an interesting phenomenon with MAE-East. Packet loss
corelates nicely to temperature.
At first I assumed the relationship was Busy Network => Hot routers
and also Busy Network => Packet Loss. But this is not the case.
It appears to be Hot Routers => Packet Loss.
Boone Boulevard MAE-East is currently running very hot. Intake temp
on our router has been up to 40 degrees today, and output at 70.
Under these conditions, the router (a 7010) starts dropping a pile
of packets occassionally. Mostly these seem to be through the AIP
and a clear int a0/0 fixes it. The time it stays fixed for is
heavilly corelated with temperature. The higher the temperature,
the shorter it stays fixed for. Eventually MFS put a fan on the
router and it seems a lot better now, intake temperature being
down to 36/37 degrees.
40 degrees is Cisco's default "warning" threshold. One would have
thought boxes should work OK at 40 degrees. On the other hand one
might also have thought a 18-22 degree aircon environment was a
prerequisite of running a decent IXP.
Is anyone else seeing high temperature and otherwise inexplicable
packet loss at MAE-East? Or does anyone else have data to corelate?
Alex Bligh
Xara Networks
X-Mailer: exmh version 2.0alpha 12/3/96
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Fri, 28 Mar 1997 20:26:57 +0000
From: "Alex.Bligh" <amb@xara.net>
Sender: owner-nanog@merit.edu
Content-Length: 1220
We've noticed an interesting phenomenon with MAE-East. Packet loss
corelates nicely to temperature.
At first I assumed the relationship was Busy Network => Hot routers
and also Busy Network => Packet Loss. But this is not the case.
It appears to be Hot Routers => Packet Loss.
Boone Boulevard MAE-East is currently running very hot. Intake temp
on our router has been up to 40 degrees today, and output at 70.
Under these conditions, the router (a 7010) starts dropping a pile
of packets occassionally. Mostly these seem to be through the AIP
and a clear int a0/0 fixes it. The time it stays fixed for is
heavilly corelated with temperature. The higher the temperature,
the shorter it stays fixed for. Eventually MFS put a fan on the
router and it seems a lot better now, intake temperature being
down to 36/37 degrees.
First I've heard of it, but sounds like you've captured a lot of
empirical evidence, so could you send me your data (hopefully
telling me whether these were output drops, input ignores, input
drops, or other errors). Did you notice any errors logged?
Anyways, send me all your usual "show ver", "show int", "show
contr cbus" info (nanog un-cc'ed, please) and I'll pass it around.
/Darren
PS - I've noticed that ever since I moved to San Diego, people
around here have been dropping like flies. Coincidence, or
cause&effect? 
70 degrees!!! This is pushing the limit for reliable operation of
electronic equipment. I know I have frequently solved problems with all
sorts of equipment (computers, modems, terminal servers, routers) by
dropping the ambient room temperature to below the level where it is
comfortable for people to work, i.e. around 15 degrees. or by adding fans
internally or externally.
I think that ambient temperatures at the tops of the racks and at the
room's outflow vents should be specified in IXP agreements along with
a protocol for dealing with this kind of problem. And most especially,
this needs to be properly engineered from the start.
Michael Dillon - Internet & ISP Consulting
Memra Software Inc. - Fax: +1-250-546-3049
http://www.memra.com - E-mail: michael@memra.com
We had a problem with this at the Atlanta-NAP, people were complaining
that it was to cold. When we warmed it up a bit, we started having
problems. So now we just tell them to dress warmly. 
Nathan Stratton President, NetRail,Inc.
I experienced an airflow problem on a 7010 once which just made it shut
down and restart when it got hot. Are you sure that this is not
happening occesionally? It would seem to me that a 7xxx router should be
able to survive its shutdown threshold without affecting performance.
Brian Horvitz
WebSecure, Inc.