From: Scott Schremmer (scotths@cs.washington.edu)
Date: Tue Nov 09 2004 - 21:56:09 PST
This paper attempts to deal with the multicasting problem of choosing a
transmission rate. Traditional attempts to adjust the rate of a multicast
have considered only one rate for all nodes. This is usually not adequate
as different nodes and links are capable of drastically different
bandwidths.
This paper introduces the protocol "Rate-driven Layered Multicast" which
is designed to deal with both static and dynamic variations in network
capacity. Other work considered this problem using a layered approach,
including packets that can be discarded. The network can drop the packets
it cannot deliver and the recipient receives a degraded version of the
stream. The difficulty with this is a lack of a method for the routers to
determine when to drop layers. The RLM scheme allows each user to
determine the number of layers to receive. Each user increases a layer at
a time until congestion becomes too much, then backs off.
The authors do a good job explaining the details and justifications for
their decisions and present simulation results. The simulation seem to
have considered a number of issues including different topologies and
scalability. Some simulation results using more realistic networks or
results on actual networks could be helpful in making this paper stronger.
This archive was generated by hypermail 2.1.6 : Tue Nov 09 2004 - 21:56:09 PST