The single collision count indicates the number of frames that were
transmitted after a single collision. The multiple collision count
indicates the number of frames that were transmitted after multiple
collisions. The late collision count indicates the number of frames that
experienced a collision after 512 bits had been sent. The excessive
collisions count indicates the number of frames that were not transmitted
because 16 collisions occurred.
If excessive collisions are happening, I wouldn't worry to much about
formulae and statistics. I'd get out there and troubleshoot the problem.
They shouldn't occur. Could you have a duplex mismatch?
Late collisions shouldn't happen either. They could mean a duplex mismatch
also.
The percent of frames that encountered a collision when transmitted versus
all frames transmitted is:
Single + multiple + late + excessive/ all frames transmitted x 100.
See more pithy comments below...
At 09:38 PM 7/25/01, Mike Fears wrote:
>Group, What is the best way to calculate collision %
>on a 10BaseT ethernet port on a Catalyst 5000/5500?
>
>Now, I have my own formula, and it is what I came up
>with after looking at CCO for the way a Catalyst 5000
>counts collisions. According to Cisco, it appears
>that:
>
>a single collision is only 1 collision (does this
>include the multi and excessive collisions?)
A frame fits into only one category. It either experienced 1 collision,
multiple (2-15), a late one, or 16 and wasn't transmitted.
>
>a multiple collision is when the same transmitted
>frame encounters more than one collision (So, if a
>frame encounters 2 or more (Or is it 2-15?)collisions
>it will increment the counter up by 1.
If it encounters more than one collision it goes in the multiple category,
but it doesn't go into the category more than once.
>and excessive collisions are collisions of more than
>16 in a row with the same tx frame.
Same as what? Excessive collisions mean there were 16 tries and then the
sender gave up.
>So, is it as simple as using the total tx frames
>(uni,multi,broadcasts) / single collisions x100?
That would be backwards for a percent of collisions compared to good frames
transmitted, wouldn't it??
>Or do you have to do something like what I came up
>with, which is
>
>single collisions + (multi collisions X 3) + excessive
>collisions X 16) + Late Collisions / Total tx frames X
>100
>
>Multiple collisions are somewhere from 2-15 and "3"
>was chosen as a guess on ports with no excessive
>collisions.
If you're looking for a reasonably exact formula, you shouldn't throw
guesses in there?? ;-) You don't need the 3. You don't need the 16 either.
>Most of the ports (about 200) that
>encountered multiple collisions did not detect any
>excessive collisions. Only about 30 saw both. On
>those, I chose "5". The good news is that I have
Once again, I don't know about this choosing of numbers to throw into the
formula. No ports should see excessive collisions. Excessive collisions are
abnormal. I would look into those if I were you.
This information is what I have gathered from reading and testing. I'm 99%
sure of its accuracy. ;-) Seriously, for questions like this, you would
have to look at the source code to be 100% sure. And, of course, there
could be differences with hardware and software versions. TAC also does a
good job describing how things work. See if there's a TAC page on this topic.
HTH
Priscilla
>mostly full-duplex ports, so I don't have to worry
>about those.
>
>
>Thanks,
>Phyrz
>CCNP
>
>
>__________________________________________________
>Do You Yahoo!?
>Make international calls for as low as $.04/minute with Yahoo! Messenger
>http://phonecard.yahoo.com/
________________________
Priscilla Oppenheimer
http://www.priscilla.com
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=13886&t=13824
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
