Verify contribution of client-server network latency in application performance

We have a typical Flash+J2EE application that makes multiple requests from client to server (over Flex remoting), which is taking quite a long time on some client systems (and hence results in poor application performance on such systems).

Now, suspecting issues with network connectivity (latency) on such开发者_开发百科 client system(s), we need to identify how far does it contribute to the slow response of the application (rather than performance issues in the application itself).

So, what are the best way(s) to diagnose this on a client system (Windows)?

Note that we have tried profiling our application, which does not indicate bottlenecks there, so we just need to clearly identify the possible network issues.

Thanks.

---

Consider the use of a network impairment solution to model these uncontrolled network characteristics for the user(s) who are complaining. The industry defacto solution set is from Shunra (http://www.shunra.com), but if all you want is single session to look at the behavior of one user, then you might consider an open source solution such as WanEm, (http://wanem.sourceforge.net).

This should allow you to model the network characteristics between a single client and the server over a congested network link to observe how the single user application performance changes with different network conditions. With WanEm you get to model one logical link between the two. With Shunra you could model your whole network if you so desired. Don't accept the default installation of WanEm on a virtual machine, timing is critical to network impairment and the clock will "float" inside of a virtual machine - Go ahead and stand up a single host for this.