Accuracy of the IP SLA UDP-Jitter measurements

[torstefan]

Dear Cisco Support Forum,

I am getting spikes in my IP SLA UDP-Jitter measurements, and is wondering if these are real spikes whcih the users would notice, or if this is just due to the router being slow to respond.

I think it might be due to the large packets and high frequency. But still, why would I get spikes ? Since the IP SLA Configuration guide says that the router uses 4 timestamps, 2 of which are on non-interupt level, just when the router sends and recieves the packet.

It is when I read this I got suspicious of the spikes in the graphs. What could be the source? Are there something iffy in my network?

Any response is appreciated.

Cheers,

Tor Stefan

Reference:

http://www.cisco.com/en/US/docs/ios/12_4/ip_sla/configuration/guide/hsoverv.html

Source:

Cisco IOS Software, C3550 Software (C3550-IPSERVICESK9-M), Version 12.2(44)SE6, RELEASE SOFTWARE (fc1)

ip sla 1

udp-jitter xx.xx.xx.191 54

request-data-size 1500

verify-data

frequency 10

ip sla schedule 1 life 12312312 start-time now

Destination:

Cisco IOS Software, s72033_rp Software (s72033_rp-ADVIPSERVICESK9_WAN-M),

cisco WS-C6509-NEB-A (R7000) processor (revision 1.2) with 983008K/65536K bytes of memory.

Processor board ID FOX10460L9F

SR71000 CPU at 600Mhz, Implementation 0x504, Rev 1.2, 512KB L2 Cache

ip sla responder

[smehrnia]

Hello Tor,

from the config you posted here, i can see you are sending a relatively large packet in a small frequency! which i dont think is necessary for udp-jitter sla, nor it is recommended.

the spikes your facing might be because of this, bcause udp-jitter is a very network intensive task. Cisco does NOT recomment using a Frequency value less than 60 sec. And depends on the purpose, reduce ur Data size, 1500 is way too much for udp-jitter !

HTH

plz Rate if it helped,

Soroush.

[torstefan]

The spikes goes away when I use 64 byte packages. (I knew that beforehand. )

And yes, I know the values are a bit high.  The point is to, in a way, "stress test" the line since there sometimes is an issue with larger packets. I monitor the CPU usage on the router and it is fine 0.03% usage. It should also be fine bandwitdth wise. (1500 bytes * 10 packets / 10 seconds) = 1.5kB/s .

I don't see why the router would be stressed by recieving 10 packets of 1500 bytes every 10th second? Why should this be an issue? The packets should also be time stamped before they reach the CPU, so it should not be an CPU issue after my understanding.

So I wonder if it truly is the IP SLA that gets taxed, or if there is something else that could be the cause?