Keepalives impact on Citrix conversation

Hi Folks,

I am having intermittent issues with Citrix XenApp Latency over WAN. I have been running wireshark on client machines during the week recording periods both good and poor performnace periods.

Firstly I have not found any errors, retransmissions, Dup acks in the captures nor have I found any 0 window size packets.

When I used the IO graph to plot the citrix TCP stream I noticed there were drop offs in packet throughput periodically during goodish periods and more frequently during bad periods. There were no errors to corealate these drops in throughput to.

I then overlayed the 'TCP.time_delta >.6' into the IO graph and noticed these values spiked before and after the citrix traffic packet throughput drops.

The only thing that appear to corealate to the packet throughout drops and the TCP time delta >.6 was that the client machies were receiving keepalive messages from 2x MS Exchange servers right before and sometimes during the Citrix packet throughput drops.

Infact in the captures taken during particularly bad periods the number of MS Exchange keepalives being received by the client is much higer than during the good periods.

I'm not sure what to make of this as normally I filter out keepalive messages I have never seen anything like this before. has anyone else?