Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 1 Jun 1999 16:36:14 -0400 (EDT)
From:      "Matthew N. Dodd" <winter@jurai.net>
To:        Dennis <dennis@etinc.com>
Cc:        Jason Thorpe <thorpej@nas.nasa.gov>, hackers@FreeBSD.ORG
Subject:   Re: xl driver for 3Com 
Message-ID:  <Pine.BSF.4.10.9906011629450.423-100000@sasami.jurai.net>
In-Reply-To: <199906011650.MAA24039@etinc.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 1 Jun 1999, Dennis wrote:
> If your nic driver chains packets (such that there is no time in
> between) you will see good throughput from the box but your overall
> network performance will suffer. A PCI card with continueous traffic
> can completely hog your lan (particularly at 10Mb/s)...which can cause
> a lot more collisions on your network as other devices will not have
> access until the hog is finished sending. For "Fairness" gaps in
> between frames are better as you approach capacity of your wire.

The ethernet spec defines the acceptable inter-frame gap.

Some cards have been known to use the minimum or less in order to 'go
faster'.  I believe that this is tunable on some cards. (LANCE comes to
mind.)  If the card isn't using the correct IFG and doesn't provide a knob
to fix it, then there isn't much we can do when the card captures the
wire.  Enabeling interrupt per packet isn't the answer either.

If you're running an unswitched LAN and use rogue cards you aren't in a
position to fuss when they do bad things.  If you cared about speed you'd
use a switch in full duplex mode.

-- 
| Matthew N. Dodd  | 78 280Z | 75 164E | 84 245DL | FreeBSD/NetBSD/Sprite/VMS |
| winter@jurai.net |      This Space For Rent     | ix86,sparc,m68k,pmax,vax  |
| http://www.jurai.net/~winter | Are you k-rad elite enough for my webpage?   |



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.10.9906011629450.423-100000>