The WinFS storage platform supports update-anywhere, peer-to-peer data replication. Due to the wide range of possible system configurations, we study the performance of the novel data replication protocol using simulation. Of interest are how many network messages are sent, the convergence time needed for a modified data object to be propagated to all sites, and how messages and convergence time are affected by failures in the network. The results for configurations of various real and synthetic network topologies show an efficient network utilization since each site receives each modification at most once despite the peer-to-peer architecture. In addition, convergence time is shown to be scalable as the number of sites increases. Finally, the protocol’s robustness to link and site failures is quantified and shown to provide good performance in the face of lost messages and transient site unavailability.