We saw comparable performance, even with 2 escon connections to host
(os/390 1.3). In addition, our escon channel is 100 pct busy. At first,
we started to use compression on all high-volume nodes. Now we are
migrating our adsm to aix. It's frustrating to have comparable network
performance in a Million$ Big Iron than in a Thousand$PC. "The next
tcpip becomes faster" - I hear this since I hear of tcpip on mvs. Maybe
the stack in 2.7 is the first usable, IBM published a webstone
benchmark.

Regards
Roger