SUMMARY: Slow network throughput

From: Evan L. Marcus [Fusion Systems Group, Ltd.] (
Date: Fri Jul 17 1992 - 15:15:16 CDT

Here is my original question...

>Here's my situation: I am using a 4/670 (4.1.2, natch) with 2 Ethernet
>controllers to route between my local LAN and a LAN in the UK. We are
>using Newbridge equipment to bridge between the UK and New York (here).
>The original configuration called for a 16kb line (out of 64kb total for
>voice, fax and data) to be designated for data. We were seeing a total
>throughput of about .15kb/sec (via ftp). That was totally unacceptable, so
>(after some analysis with nfswatch, traffic, netstat, and perfmeter) we
>called in the Newbridge tech to look at their equipment. He found
>that just about all LAN traffic from New York was being routed onto the UK
>net, and that this may have accounted for our slow throughput.
>We were able to borrow the voice lines for awhile, and upped the data line
>to 48kb, and throughput increased from .15kb/s to 3.4kb/s. A 3x increase
>in bandwidth gave us a 22x increase in throughput. The obvious conclusion
>is that there is some fixed amount of overhead.
>The question is: why would we be seeing traffic from one LAN (192.9.203)
>routed onto the other LAN (192.9.205)? I tried manual routing, and automated
>routing (in.routed) to no avail.
>The other question is: should there be this level of overhead?
>Finally, I have heard about patch 100283-02, for routed. The archie server
>at rutgers doesn't know about it. If someone has it, and could email it to
>me...I'd be terribly grateful. I'm not sure it'd actually help, but it
>won't hurt, so what the hey?

After I posted this, I used etherfind to analyze the network, and it turns
out that the guy from Newbridge was wrong; there was NO traffic whatsover
on the UK link that shouldn't have been there. The only traffic on that
net that was not from or to a machine the .205 subnet was the traffic
directly between the two bridges.

Nevertheless, we still took the opportunity to get our hands on Sun patch
100343-02, which improves in.routed. I installed it, but it's not really
clear if it matters or not.

As it stands now, we are still having problems with the line. We have
gotten some more bandwidth, and whenever we do, the throughput improves
much more than linearly.

But, as this problem is solved about as well as it's going to be, I thought
I would stop sitting on the summary, and post what I have and have

Thanks to:

for various suggestions, and of course for all mentioning the near
impossibility of the Sun mis-routing packets!

Thanks to all...

WHO: Evan L. Marcus		  || There is no truth to the rumor that
WHAT: Fusion Services Group, Ltd. || Bill and Hillary Clinton plan to 
WHERE: New York, New York, USA	  || name their next children Soho and
HOW:		  || Little Italy.

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:45 CDT