Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 30 Nov 1995 13:52:27 -0800
From:      David Greenman <davidg@Root.COM>
To:        "Garrett A. Wollman" <wollman@lcs.mit.edu>
Cc:        gpalmer@westhill.cdrom.com, questions@freebsd.org
Subject:   Re: IPX status and routing 
Message-ID:  <199511302152.NAA02372@corbin.Root.COM>
In-Reply-To: Your message of "Thu, 30 Nov 95 10:34:49 EST." <9511301534.AA08546@halloran-eldar.lcs.mit.edu> 

next in thread | previous in thread | raw e-mail | index | archive | help
><<On Wed, 29 Nov 1995 18:33:21 -0800, gpalmer@westhill.cdrom.com said:
>
>> 2) I doubt it. Dedicated routers can do slightly faster packet
>>    switching as they start routing when they've recieved the header,
>>    whereas BSD has to wait for the entire packet.
>
>A good PCI machine (like a 100-MHz Dell I have in the lab) with good
>PCI interfaces (like the DE500s we have in that machine) can forward
>about 15,000 minimally-sized packets per second (all to the same
>destination, but my tests indicate that this shouldn't matter).  That
>comes to about 7.7 Mbit/s.  Not as fast as we would like, but well
>within the envelope for most uses.  (This load is from an input stream
>of 18,000 packets per second over Fast Ethernet.)  For larger packets,
>the rate goes down somewhat.

   The maximum packet rate given overhead, collisions, and minimum ethernet
packet size (60 bytes) is about 12,500 packets/second. ...so 15,000 packets/
second is higher than what normal ethernet can do for one interface. On the
other hand, fast ethernet should be about 10 times that...so we have a long
way to go to get to 125,000 packets/second. :-)

-DG



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199511302152.NAA02372>