Measured characteristics of process interarrival times across a LAN

Our testbed consists of two PCs on a multiple PC-LAN. The two PCs are engaged in syn chronous interprocess communication, otherwise known as client-server communication. Commu nicating processes are continually being created at both client and server. We are interested in their intercreation time distribution-which is really the interarrival time distribution-for the pur pose of modeling the client-server communication at any layer of the network architecture. What is the effect on the intercreation time distribution of a LAN bridge being inserted between client and server? What is the effect of varied network loading? When is there autocorrelation in the distribution of intercreation times and when is the distribution most suitable as an input to a performance model?