Re: [nsp] bit rate accuracy?

From: Gert Doering (gert@greenie.muc.de)
Date: Tue Nov 13 2001 - 14:45:13 EST


Hi,

On Tue, Nov 13, 2001 at 12:56:51PM -0500, Jared Mauch wrote:
> Are you using any fancy features (Rate-limit, etc..?) on the interface
> in question? There was one bug I saw in IOS whereby it would report
> 2+Gb/s on any interface with rate-limit configured amongst other
> issues.

I've seen this on the "show int" 5 minute output average values in
12.0(about 14)S, without any fancy queueing going on. It happens every
time the output byte counter wraps over 2Gbyte - this seems to inject
a "4 Gbyte in a nanosecond" spike into the decaying average...

> I'm not aware of any of those existing in current software
> versions but they did exist in the past (somewhere around 12.0.14S->12.0.17S)

Current versions of 12.0S have funny bugs with "show int" counters not being
updated unless you do a "show int acc" before - happens on 12.0(17)S and
later if the interface has wrapped often enough (but not easy to reproduce).

Does not happen for SNMP counters.

> The interface may also be taking into account 95% type information
> as far as bit rates, which mrtg may not pick up any 2-5 second bursts but
> the statistics collector within IOS may pick up also. I suspect that is
> what is really going on.

Actually from my experience, very bursty traffic is not properly averaged
within IOS, so the average based on counters is *higher*. I haven't seen
the other way round (counters lower than average) for extended periods of
time lately - it could happen "for a few samples" if the traffic suddenly
drops, like "20 minutes 100 Mbit, then 20 minutes only 10 Mbit" - counter
statistics get this immediately, 5 minute average decays slower...

gert

-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert@greenie.muc.de
fax: +49-89-35655025                        gert.doering@physik.tu-muenchen.de



This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:23 EDT