Nick, thanks for reply, I'm definitely seeing the difference, I will do more tests to figure our the bottleneck, have to think a bit about how to measure the delay exactly too (currently I'm using SecureCRT as SSH client, should probably try something else)