"What is the sound, Grasshopper, of one hand clapping?" said the traditional master. And the techie student answered, "Start an audio recorder. Clap one hand. Play back the recording. That is the sound of one hand clapping." If there are some critical applications such as ERP, presumably there are at least LAN-based testbeds for them. Why not take some actual measurements, especially that show the interactions of multiple clients and servers? I'd try to do this as much as possible on 10 Mbps switched Ethernet, to be at least somewhat comparable to WAN links. I suspect you are looking more for utilization than absolute lengths, but you should be able to do some extrapolation. As a general observation in queueing theory, latency starts going up, with reliable transfer protocols, much above 50% utilization. At a given MTU level, I suspect the data link header lengths will not be a significant difference. Signaling? FR and ATM have them, but LMI traffic isn't all that frequent. Will you be running CDP? Signaling involved in setting up VCs only occur at setup (surprise!), and whether they are Q.2931 or PPP, will be lost in the noise if the circuit stays up for any appreciable time. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=10058&t=10058 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]