[c-nsp] IP SLA monitoring

[Jee Kay]

On 22/11/06, Rodney Dunn <rodunn at cisco.com> wrote:
> But can't you set the frequence down to 1 second under
> the probe?

Quite possibly... I don't know what the effect of having so many
skipped operations might be.

On a side note, I'm very confused by the numbers being produced by
these jitter graphs. I have a Cat4500 SupV plugged directly into a
Cat4500 SupV via an FE copper cable, and am getting numbers like these
coming out:

RTT Values (milliseconds):
NumOfRTT: 1000  RTTAvg: 2       RTTMin: 1       RTTMax: 23
RTTSum: 2656    RTTSum2: 8720

Jitter Values (milliseconds):
MinOfPositivesSD: 1     MaxOfPositivesSD: 20
NumOfPositivesSD: 77    SumOfPositivesSD: 265   Sum2PositivesSD: 1553
MinOfNegativesSD: 1     MaxOfNegativesSD: 20
NumOfNegativesSD: 102   SumOfNegativesSD: 267   Sum2NegativesSD: 1457
MinOfPositivesDS: 1     MaxOfPositivesDS: 4
NumOfPositivesDS: 148   SumOfPositivesDS: 215   Sum2PositivesDS: 437
MinOfNegativesDS: 1     MaxOfNegativesDS: 4
NumOfNegativesDS: 124   SumOfNegativesDS: 217   Sum2NegativesDS: 543


Both have 'rtr responder' configured. How on earth is RTTmax 23ms?!
And how am I managing to get jitter of 5ms+ on such a direct link? It
seems to me that the responder is not responding very quickly..

On another link, between a c2621 and the same Cat4500/SupV, with the
monitor configured on the c2621, I consistantly see the jitter 'from D
to S' (ie 4500->2621) being a rock solid 4ms, but the jitter from S to
D varies massively between 2ms and 20ms. Again these two devices are
connected by nothing more complicated than a piece of copper cabling.

Am I doing something drastically wrong, or should I expect the
responder to take significant amounts of time to respond? I have some
30-40ms latency links that show RTTmax at over 900ms occasionally,
presumably again due to the same problem.

Thanks,
Ras