Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 19 Jun 1997 11:07:59 +0200
From:      sthaug@nethelp.no
To:        tom@sdf.com
Cc:        ccsanady@scl.ameslab.gov, hackers@FreeBSD.ORG, matt@3am-software.com
Subject:   Re: Network concurrency problems!?
Message-ID:  <7618.866711279@verdi.nethelp.no>
In-Reply-To: Your message of "Wed, 18 Jun 1997 18:16:49 -0700 (PDT)"
References:  <Pine.BSF.3.95q.970618181004.11965A-100000@misery.sdf.com>

next in thread | previous in thread | raw e-mail | index | archive | help
>   Would a TCP socket have more system overhead than a file on a UFS
> filesystem?  I guess you can tess this by running dd on a mfs, and
> tcpblast/netpipe/ttcp on loopback?  Somehow I think the socket would be
> faster.

I don't know enough about the file system to say whether this is a
relevant comparison on one host.

It's not a very good comparison to the Ethernet case for the simple
reason that the Ethernet MTU is 1500, while the default loopback MTU
is 16384.

I did a couple of quick tests here on a PPro-200. With the default
loopback MTU, I consistently got around 44 MByte/s (M=1048576) using
ttcp. With a 1500 byte MTU, I'm down to 27 Mbyte/s - a significant
difference.

Using the loopback interface the packet needs to traverse both down
and up the protocol stack on the same host, so this would tend to
decrease performance. I don't know whether the increased performance
from a large loopback MTU offsets the decreased performance from
traversing the protocol stack twice.

Steinar Haug, Nethelp consulting, sthaug@nethelp.no



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?7618.866711279>