*BSD News Article 22157


Return to BSD News archive

Xref: sserve comp.os.386bsd.misc:1192 comp.os.linux:56027
Newsgroups: comp.os.386bsd.misc,comp.os.linux
Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!elroy.jpl.nasa.gov!swrinde!cs.utexas.edu!asuvax!chnews!ornews.intel.com!percy!agora!davidg
From: davidg@agora.rain.com (David Greenman)
Subject: Re: NetBSD TCP/IP network benchmarks
Message-ID: <CEnKBv.Fx5@agora.rain.com>
Organization: Open Communications Forum
Date: Sat, 9 Oct 1993 23:09:30 GMT
Lines: 32

>Measurements were done using 2 33 mhz 486dx boxes with 16 MB RAM and
>3c503 ethernet cards. The machines were run in single user mode on a

>TCP throughput was measured using ttcp (anon ftp from sgi.com in
>sgi/src/ttcp) which is a 16MB one-way memory-to-memory transfer.
>
>system			codebase	kbytes/s
>
>NetBSD 0.8		BNR2		320		

   Well, NetBSD 0.8  has the same broken and inefficient ethernet drivers as
386BSD 0.1 did. NetBSD, FreeBSD, and even Bill's up-and-coming 386BSD 0.2
doesn't use them anymore - they all use the 'ed' ether driver that I wrote.
Here's why:

[corbin::davidg]$ ttcp -t implode
ttcp-t: buflen=8192, nbuf=2048, align=16384/+0, port=5001  tcp  -> implode
ttcp-t: socket
ttcp-t: connect
ttcp-t: 16777216 bytes in 15.10 real seconds = 1085.16 KB/sec +++
ttcp-t: 2048 I/O calls, msec/call = 7.55, calls/sec = 135.65
ttcp-t: 0.0user 6.3sys 0:15real 42% 0i+0d 0maxrss 0+0pf 5677+8csw
0.060u 6.400s 0:15.19 42.5% 0+0k 0+0io 0pf+0w

   Perhaps what's even more important than the 1085k/second is the fact
that it only takes 42% of the CPU. This is with FreeBSD-1.0-Epsilon
w/rev. 2.6 of my ed ethernet driver using two SMC 8013EPC ethernet adapters
between a 486DX2/66 and a 486DX/33. The results are about 1% lower between
two 486DX/33's. The results are about 10% lower (measured by me at 985K/sec)
when using 16bit 3c503 adapters.

-DG