Analyzing poor file transfer speed issues
I have an issue where users trying to down a file via load balancer get just 1/3 of the throughput of when they go directly to the pool servers behind the load balancer. Surprisingly the pcap via the LB actually had larger window sizes than the caps for the the xfer that bypassed it. In fact when going to the LB the window size was over 1MB and when bypassing it the window size was a miniscule 766B per packet. I was expecting just the opposite. Before I call the LB tech support people, I wondered if any Sharkers here might have some more things to look at to help me analyze what's happening with the file transfer. The capture is happening at the client side of things.
For Sharkers to be able to help you they need to look at the trace file. Also with such cases as yours it's better to have traces from both LB sides.
As for Windows size of 766 Bytes - have you checked Window scaling factor? You can spot it only in SYN packets.
Window size of 766Bytes is kinda strange. And what do you mean by "window size per packet"?