Measuring Load Speed on Thin Client

[Balkoth]

I'm trying to do something a bit unusual here and was looking for some ideas. Specifically, I'm trying to measure the relative performance of different encoding schemes across different network speeds and packet loss. One of the main ideas of measurement I have is measuring the time it takes for the following to happen:

1. Press enter on the thin client which sends data to the server and the server starts an MRI viewing application
2. Server then does all the processing work and sends the screen image data back in a preselected encoding scheme
3. The client "unpacks" the data and fully loads the image

So basically, the time it takes from the time you hit "go" to the time the image is loaded on the *client* side.

I have some ideas I'm looking into with for the "go" timestamp, ranging from just trying to timestamp on the client or timestamping when the server receives the message (and subtract off the latency and compensate for different internal times potentially), but I'm feeling stumped about how to track when the MRI image "loads" (I say "loads" because it's not a static image but rather a rotating display that can be manipulated).

Thoughts on how to do that? Or alternative measures of performance?

Thanks!

[Bonji]

I had an initial reply all typed out that I wasn't happy with, so I deleted it.

You could use a program like NetLimiter Lite to measure the actual network bandwidth used by the VNCViewer process, however you won't see client impact measured this way (primarily CPU and actual visual quality/performance).

For timings, you could script out what you want done on the server's end, and just use an old-fashioned timer to see how long it takes to process on the viewer side. Consistency can be hard to get here as the application itself needs to do everything at the same speed each time the process is performed on it.