Abstract: The Selectively Reliable Multicast Protocol (SRMP) provides an approach to reliable multicast that is specialized to distributed virtual simulation. SRMP operates in three modes of reliability: best-effort transmission of transient data that does not need reliable transport; reliable transmission of object state data; and timely, reliable transport of data to a single, dynamically determined receiver within the multicast group. This paper deals with the first two modes, which dominate SRMP network performance. Any multicast protocol, when used across wide area networks to share information among thousands of objects in a real time simulation, will create traffic flows with many complexities. In order to evaluate the merits of using SRMP, a mechanism to measure its performance across a network is necessary. This paper presents an approach to measure the bounds of performance of the SRMP in real time simulation environments. The approach is based on a Markovian model of object activity in the simulated environment, and is implemented in the context of a cluster of workstations that emulate wide-area network performance using our protocol. We present initial results of our SRMP evaluation in this environment.