IP SLA average response time calculation formula

[Ferdi0802]

What is the formula to calculate average response time when using IP SLA type ICMP echo operation?

[Joe Clarke]

For all successful probes, this is the mathematical mean of the samples.

(X1 + X2 + X3 ... + Xn) / n = avg response time

Where X is the RTT of each probe.

[Ferdi0802]

Let say I want a network management system measures the average RTT.  The router will send ping every 30 seconds and NMS will poll the data from that router every 5 minutes. From formula you mention above, which OID should be used to read the history from X(1) until X(n-1)? Using "sh ip sla monitor statistics"  I only see latest RTT values.

[Joe Clarke]

Sorry, I was confusing jitter with ICMP echo.  The jitter operation tracks averages.  ICMP echo tracks the latest RTT value.  You can enable history to view the RTT value from previous collector executions, but that will still track the latest RTT (in rttMonHistoryCollectionCompletionTime).  Any avergae would need to be calculated externally.

[Ferdi0802]

So, if my NMS poll data every 5 minutes and I configure ICMP echo to operate every 30 seconds, will NMS only collects 2 RTT values (one from rttMonHistoryCollectionCompletionTime and one from rttMonLatestRttOperCompletionTime)?

[Joe Clarke]

Well, rttMonHistoryCollectionCompletionTime could have multiple records per collector depending on the sample time and bucket size.  But there will be one occurrence of rttMonLatestRttOperCompletionTime per collector.

[Ferdi0802]

Here is my configuration

ip sla monitor 100
type echo protocol ipIcmpEcho x source-ipaddr y
request-data-size 100
timeout 2000
tag TES-PING
frequency 30
buckets-of-history-kept 5

I use MIB browser to see rttMonHistoryCollectionCompletionTime and it gives me nothing. But I can get rttMonLatestRttOperCompletionTime value, so there is no problem with my MIB browser.

What configuration do I miss?

[Joe Clarke · Accepted solution]

You need to add:

filter-for-history all

lives-of-history-kept 2

[Ferdi0802]

Thanks, it works. It's a little out of topic, but, how can I measure packet loss during that period? I only see "number of packet failure" field, and I think the value will be accumulated, so I can't rely on that parameter.

[Joe Clarke · Accepted solution]

If you want packet loss, you should look at using UDP jitter and an IP SLA responder.  The echo operation will just show you RTT.