Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!harpo!seismo!hao!hplabs!sri-unix!walsh@bbn-labs-b
From: walsh%bbn-labs-b@sri-unix.UUCP
Newsgroups: net.unix-wizards
Subject: Trailer protocol on 4.2BSD
Message-ID: <131@sri-arpa.UUCP>
Date: Tue, 24-Apr-84 07:14:22 EDT
Article-I.D.: sri-arpa.131
Posted: Tue Apr 24 07:14:22 1984
Date-Received: Mon, 30-Apr-84 05:46:16 EDT
Lines: 23

From:  Bob Walsh <walsh@bbn-labs-b>


I would like to find out if anyone has used trailers and found them
beneficial.  My own tests have shown that they have little effect on
throughput.

My tests used a VAX 780 and a VAX 750 connected to an ethernet
by Interlan controllers.  I have tried two programs. The first is my own
benchmarking program to measure bit rates.  The other is ftp.  I used the
benchmarking program to measure bit rates when writing 1, 2, 4, 8, 16, 32,
64, 128, 256, 512, 1024, and 2048 byte sized buffers onto a TCP connection.
For the different sized buffers, sometimes trailers produced higher bit rates
and sometimes trailers produced lower bit rates.  Trailer throughput was
always within 3% the throughput achieved without trailers.  Similar results
were found with ftp.  During these tests, socket send and receive
buffering high water marks were set at 4096 bytes.

Does a different hardware configuration bring forth the 200Kb/sec gain
mentioned in the "Networking Implementation" paper found in volume 2C?
Yes, I had remembered to turn trailers on for both VAXes.

bob walsh