Hi. In diagnosing application performance problems, I am very often faced with transactions that are slow because they are "chatty" - that is, requiring a large number of round trips to complete the transaction. (Of course, the impact of the chattiness is also dependent on the end-to-end latency between the endpoints.) In the past, I have used another diagnostic tool - Opnet. This tool does a good job of reporting on the chattiness / round trip count, and its precise contribution to the transaction duration. But, Opnet is an expensive tool, not available to many of us packeteers. So my question is... What features and/or techniques are recommended for measuring the impact of chattiness with Wireshark? One technique, I suppose (thinking out loud now), would be to filter on packets between the 2 endpoints, and exclude those non-data-bearing packets, and then count the number of packets and the elapsed time between the first and last packets. This should give you number of round trips (half the packet count), and the duration of the transaction. This would give you an IDEA of the impact of chattiness, but the elapsed time would also include time spent on either end processing between packet exchanges, so you would have to be careful how to interpret this. Anyway, does anyone have other suggestions? Or, is there a feature/option of Wireshark that I have ignored? Thx all, feenyman99 asked 04 Jul '11, 07:30  feenyman99 |
One Answer:
I think you can get a good idea by looking at the conversation summary. Just go to statistics - conversations and click the appropriate tab (usually TCP) you can then gat an idea of duration, avg packet size, etc. Hope it helps! answered 06 Jul '11, 13:32  ericinsd |
