5 Replies Latest reply: Oct 5, 2009 11:56 AM by byrona RSS

Cisco IP SLA Latency Behavior

byrona

We have several Cisco devices running IP SLA and I have setup some data-collections using the UnDP.  What I have found odd is the Source to Destination is almost an exact mirror of Destination to Source and this seems odd to me.

I am using Transform Results formulas as follows...

Latency Destination to Source:  {rttMonLatestIcmpJitterOWSumDS}/{rttMonLatestIcmpJitterNumRTT}

Latency Source to Destination:  {rttMonLatestIcmpJitterOWSumSD}/{rttMonLatestIcmpJitterNumRTT}

Below is a screenshot of the graphs for one of the devices but all of the devices are pretty much similar.

I would be very interested in hearing from others that have some experience with this Cisco IP SLA stuff to find out if this is expected behavior or not.

Thanks in advance for any pointers!

 

 
  • Re: Cisco IP SLA Latency Behavior
    derhally

    You may try:

    Latency Destination to Source:  {rttMonLatestIcmpJitterOWSumDS}/{rttMonLatestIcmpJitterNumOW}

    Latency Source to Destination:  {rttMonLatestIcmpJitterOWSumSD}/{rttMonLatestIcmpJitterNumOW}


    rttMonLatestIcmpJitterNumOW is the number of successful one way trip time measurements where as

    where as rttMonLatestIcmpJitterNumRTT is the total number of RTT's that were successfully.

    This probably won't fix the issue where you see a mirror image of the latency.

    What version of the IOS are you using?

  • Re: Cisco IP SLA Latency Behavior
    derhally

    The other question I forgot to ask, what do you see when when you look at the statistics of the operation through the IOS CLI?

    You can look at the statistics by running the command

    sh ip sla statistics <operation number>

    or if you have an older version IP SLA

    sh ip sla monitor statistics <operation number>

    • Re: Cisco IP SLA Latency Behavior
      byrona

      Thanks for the info, I think the divisor you suggested is better, while it yields the same numbers it is the more accurate data point to be using so I updated my collection.

      The data that I am getting accurately reflects what I am getting when I look at the data via CLI.

      It just seemed odd that while not an exact mirror, the data would have a mirror like look to it, especially for the spikes.  It's almost like it can tell that it took too long one way so it tries to make up for it on the other direction.

    • Re: Cisco IP SLA Latency Behavior
      PinkieT

      We are using "sho ip sla stat 1" to render the data via the CLI

       

      Thanks for the help!