Finding latency

[mohan]

I have some latency issue in my n/w.

If i am pinging from one switch of the cluster-X, i am getting the result satisfactory.However,If i am pinging from switch 2 of cluster -X to same server i am getting high latency.

If i ping from user workstation,i am getting 10 to 13 packets as normal & next 3 packets round time trip very high.
i am not getting any packet drops..

Has anyone encountered this problem..

[Michael McNamara]

Hi abcd,

You really haven't provided very much information... so I'm not sure we'll be able to provide any help.

In addition, with a display name of abcd I'm not sure how you expect people to take time out of their busy scheduled to answer your question.  Perhaps you should adjust your userID or display name and then re-post your question with some specific details.

Cheers!

[mohan]

hi michael,

request your apology for the above mentioned id.

i have 4 distribution cluster's of 8600's connected to core cluster(8600's) through rsmlt between Distribution & core.
Server farm is connected to core.
the above mentioned issue is if i ping to a server in server farm,from Distribution 3 sw#1 then i get round time trip average 0.7ms ,if i ping to same server from Distribution 3 sw#2,i am getting 1.8ms for each packet & after 10 to 13 packets,14,15,16th packets have round time trip 38ms.
the same pattern is observed if i am pinging from user PC.

Access switches are connected through smlt to distribution.The user Pc's and the server are in same VRF.No problem reported from users connected to other distribution accessing same application.

[Michael McNamara]

Hi Mohan,

I'm sorry if my reply seemed overboard... just trying to keep the posts relevant and the discussion meaningful.

That's an interesting problem... the traffic to/from a specific source and destination should take the same path that's the whole point behind MLT. Unfortunately ICMP is the lowest priority traffic and is usually the first to be delayed/dropped. What if you connect a PC to the same switch as the server farm, do you get the same results?

You could check the CPU/SF utilization of your ERS 8600 switches with the following command show sys perf, run it a few times over a 60 second period to see what the average is and if there are any spikes.

I would suggest you try and isolate either the network or the server farm itself, that will help reduce the number of variables and give you a hint at where to concentrate additional testing.

Good Luck!