This paper evaluates the congestion control performance of Pseudofed, a congestion-controlled, reliable multicast transport protocol for bulk data transfer. Pseudofed's congestion control mechanism is based on the concept of representatives, a small, dynamic set of multicast group members. By reducing the congestion control problem to a bounded set of receivers, representatives allow the point-to-point congestion control model used by unicast protocols like TCP to scale to larger multicast groups. Other features that contribute to the scalability of Pseudofed's congestion control algorithm are: (1) attempting to distinguish between correlated and uncorrelated packet losses, (2) not requiring complete knowledge of the multicast group, and (3) not exchanging control communication with congestion free subtrees.

This paper evaluates a congestion control mechanism for reliable multicast applications that uses a small dynamic set of group members, or representatives, to provide timely and accurate feedback on behalf of congested subtrees of a multicast distribution tree. Through simulations, we evaluate our algorithm with and without TCP cross traffic. This study demonstrates that while our algorithm makes use of bandwidth when available, it responds to congestion in a timely fashion thereby limiting data drops. When competing for bandwidth with TCP traffic, our algorithm gracefully relinquishes bandwidth. While not as aggressive as TCP, it does grab a reasonable portion of the bandwidth yet keeps the overall multicast packet loss low.

For a reliable, feedback dependent multicast transport protocol to scale, it must avoid the feedback implosion problem, particularly if the protocol targets arbitrarily large multicast groups communicating over lossy networks. Most existing suppression based feedback control mechanisms address the implosion problem using timers based on round-trip time (RTT) estimates between each receiver and the source. The algorithm presented has three major benefits: it does not need to compute the RTT from all receivers to the source, does not require knowledge of group membership, and provides prompt feedback. A small set of representative receivers and probabilistic suppression are used to limit feedback. We believe that this approach will perform well in real networks. Simulations show that for various multicast group sizes, a few representatives can keep the amount of feedback low while not degrading feedback timeliness.

This paper reports the design, implementation, and performance of a scalable and efficient tool to replicate Internet information services. Our tool targets replication degrees of tens of thousands of weakly-consistent replicas scattered throughout the Internet's thousands of autonomously administered domains. The main goal of our replication tool is to make existing replication algorithms scale in today's exponentially-growing, autonomously-managed internetworks.