[nsp] single-col, multi-col, 6509

[Church, Chuck]

There is a close relationship between collision rate and link utilization in half duplex ethernet.  The Cisco design guides I've seen in the past mention 0.1% as the threshold for increasing speed or moving to full duplex.  Personally, I think that 0.1% rule is a little overkill, so I've gone with 1%.  If you've got 20% collisions, you're forced to resend 1 out of every 5 frames.  You'll certainly be seeing multiple and excessive collisions at that rate, leading to retransmits at layer 3/4, which leads to yet more traffic.  20 to 50% is WAY too much traffic for a half duplex segment, especially if there are more than 2 devices on that collision domain.

Chuck Church
Lead Design Engineer
CCIE #8776, MCNE, MCSE
Wam!Net Government Services - Design & Implementation Team
13665 Dulles Technology Dr. Ste 250
Herndon, VA 20171
Office: 703-480-2569
Cell: 703-819-3495
cchurch at wamnetgov.com
PGP key: http://pgp.mit.edu:11371/pks/lookup?op=index&search=cchurch%40wamnetgov.com

> -----Original Message-----
> From: sthaug at nethelp.no [mailto:sthaug at nethelp.no]
> Sent: Monday, March 22, 2004 4:14 PM
> To: Church, Chuck
> Cc: cisco-nsp at puck.nether.net
> Subject: RE: [nsp] single-col, multi-col, 6509
> 
> 
> > Compare the number of single collisions to the total number of
> > packets output.  If you're getting more than 1 collision per 100
> > packets output, it's probably overloaded.
> 
> Sorry, collision rate doesn't work that way. If you have a half
> duplex segment, collisions are expected - and 20% or 50% could
> be just fine. It all depends on the total amount of traffic you
> have on your half duplex segment - *this* is the interesting
> number, not the collision rate.
> 
> > Like the others
> > mentioned, check for any duplex mismatches, or FCS errors on either
> > side.
> 
> Good points.
> 
> Steinar Haug, Nethelp consulting, sthaug at nethelp.no
>