Topology Sensitive Congestion Control
for Real-Time Multicast
Srinivasan Jagannathan, Kevin C. Almeroth
& Anurag Acharya
University of California Santa Barbara
Abstract:
The multicast-based distribution of streaming media is
conjectured to be an important component of future net-work
services. Congestion control has emerged as a major
hurdle in the large-scale deployment of such services.
Heterogeneity in transmission media, end-systems, and traffic
flows have significantly increased the complexity of the
problem. Researchers have sought to address the problem
through such mechanisms as layered encoding and TCP-style
formulae. The end-to-end paradigm has been the guiding
principle for these approaches. In parallel, many tools
are being developed that provide snapshots of network internals.
Of interest to us are tools that accurately construct
or estimate the topology of a multicast tree. In this paper, we
explore whether these tools can assist in providing congestion
control for real-time multicast. Through this exercise
we find that the multicast tree topology, if available, can be
used to regulate the flow of streams. In this paper, we seek to
understand the additional benefits of using this information
by focusing on the extreme case where the complete tree
topology is known. We develop an algorithm which uses
this information and layered streams to provide robust congestion
control. We evaluate our algorithm using ns, the
network simulator. Our results show that our algorithm is
robust and converges to "fair" subscription levels for each
receiver.