[c-nsp] SNMP MIB update interval on CISCO?

Dale W. Carder dwcarder at wisc.edu
Thu Mar 27 11:45:05 EDT 2008


On Mar 26, 2008, at 12:58 PM, Gurung, Provin wrote:
> Thanks for the information. Do the routers maintain a timestamp of  
> when
> they last updated their MIB. The time difference between the updates
> will give me a good indication of the average traffic being  
> observed by
> the router.

I don't think so, and each ifindex could possibly update at
a different time.

I've tried sampling at 5 min, 1 min, 30 sec, 10 sec and have
definitely seen the accuracy issues outlined below.

Dale

> --Original Message-----
> From: cisco-nsp-bounces at puck.nether.net
> [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of mack
> Sent: Tuesday, March 25, 2008 5:42 PM
> To: cisco-nsp at puck.nether.net
> Subject: Re: [c-nsp] SNMP MIB update interval on CISCO?
>
> On the 6500/7600 platform you can set:
>
> service counters max age 5
>
> Obviously this is platform specific.
> On high CPU load SNMP counters may be updated at a longer interval.
> At extreme levels SNMP may stop updating completely until the load  
> goes
> down.
>
> Most platforms have a best-effort interval of 10 seconds.
>
> As for the accuracy:
>
> % Accuracy~=update interval*100/sample interval
>
> So with a 60 second sample interval and a ten second update interval
> you can have a variance of ~17%.
>
> With a 5 minute sample interval the variance is ~3%.
>
> The accuracy is an approximation as the update interval is best- 
> effort.
>
> To sum it up you can sample slower and get more accurate readings or
> faster and
> get less accurate readings.
>
> --
> LR Mack McBride
> Network Administrator
> Alpha Red, Inc.
>
>
>> Message: 9
>> Date: Tue, 25 Mar 2008 20:02:23 -0000
>> From: "Dean Smith" <dean at eatworms.org.uk>
>> Subject: Re: [c-nsp] SNMP MIB update interval on CISCO?
>> To: <cisco-nsp at puck.nether.net>
>> Message-ID: <005401c88eb3$29605b50$01011bac at thor>
>> Content-Type: text/plain;       charset="us-ascii"
>>
>> Basically - No.
>>
>> Counters like those are really only valid for polling every minute at
> a
>> minimum and even then you can occasionally see strange affects due to
>> the
>> internal updates happening out of sync.
>>
>> If you essentially need granularity of a second or so sampling you
>> probably
>> need an external tool.
>>
>> Dean


More information about the cisco-nsp mailing list