[eepro100] eepro100 and large packets

Dan Stromberg strombrg@nis.acs.uci.edu
Tue, 31 Oct 2000 14:07:22 -0800


--ugCSa1rMkiCAmd9S
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

For what it's worth:

sabaki-strombrg> uname -r
2.2.14-5.0

sabaki-strombrg> ping hydra.acs -s 65000 -c 10
PING hydra.acs.uci.edu (128.200.16.3) from 128.200.34.80 : 65000(65028) byt=
es of data.
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D0 ttl=3D255 t=
ime=3D122.4 ms
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D1 ttl=3D255 t=
ime=3D117.0 ms
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D2 ttl=3D255 t=
ime=3D117.4 ms
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D3 ttl=3D255 t=
ime=3D117.7 ms
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D4 ttl=3D255 t=
ime=3D117.2 ms
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D5 ttl=3D255 t=
ime=3D117.4 ms
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D6 ttl=3D255 t=
ime=3D117.3 ms
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D7 ttl=3D255 t=
ime=3D117.1 ms
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D8 ttl=3D255 t=
ime=3D117.4 ms
65008 bytes from hydra.acs.uci.edu (128.200.16.3): icmp_seq=3D9 ttl=3D255 t=
ime=3D117.4 ms

--- hydra.acs.uci.edu ping statistics ---
10 packets transmitted, 10 packets received, 0% packet loss
round-trip min/avg/max =3D 117.0/117.8/122.4 ms

sabaki-strombrg> cat /etc/conf.modules
alias eth0 eepro100
alias parport_lowlevel parport_pc
alias sound-slot-0 esssolo1

sabaki-strombrg>=20


I saw the same good results on a 2.2.16 machine with an eepro100 card.

I'm no networking expert, but I might wonder if there weren't a bad
switch/hub/router or something involved.

On Tue, Oct 31, 2000 at 11:14:58AM -0500, Philip E. Varner wrote:
> Where I work, our newest servers have eepro100 NICs.  They're running
> kernel 2.2.14.  Our primary application uses packets 65000 bytes in length
> for communication.  However, when sending large packets like this,
> performance plummets at a packet size of around 64,590 bytes.  Below this,
> I get minimal packet loss, but above this it jumps to around 50-70%
> _loss_.
>=20
> I've tried this on other machines (de4x5 driven) on the same switch and it
> doesn't happen.  Has anybody seen this before?
>=20
> Phil Varner
>=20
>=20
> --
> "We have fewer and fewer excuses for ignoring or being indifferent or=20
> sloganizing the very serious and in many ways growing injustices=20
> in our society." -- Ralph Nader in his speech to the NAACP
>=20
>=20
>=20
> _______________________________________________
> eepro100 mailing list
> eepro100@scyld.com
> http://www.scyld.com/mailman/listinfo/eepro100

--=20
Dan Stromberg                                               UCI/NACS/DCS

--ugCSa1rMkiCAmd9S
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.4 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE5/0KZo0feVm00f/8RAto1AJ95WOhMyORL5JOCeQeh6Ll64YT6xQCeKg52
Z0FcXB9j2+8EQUbJtS2zChs=
=eApT
-----END PGP SIGNATURE-----

--ugCSa1rMkiCAmd9S--