Review: Receiver Driven Layered Multicast

In this paper author has presented a transport protocol to deliver
multicast contents in a way that maximizes individual links’ utilization.
The protocol combines the layered source coding algorithm with a layered
transmission scheme. The protocol, called Receiver Driven Layered
Multicast (RLM), requires no active participation from the sender except
that the sender encodes the a signal in multiple layers with each layer
providing fine grained refinement to the layer below. At each intermediate
hop only selective layers, starting from bottom and moving upwards, are
forwarded based on the downstream link capacity thus taking care of
heterogeneity of network link bandwidth.

RLM, as mentioned earlier, does not require sender to decide the data rate
for receivers, despite each receiver decides it own data rate at which it
can efficiently receive the signal. The sender transmits each layer of
signal on a separate multicast group and the receivers join and leave
these groups. In order to maximize the use of available bandwidth,
receiver requests to join higher layer group after specified interval
(join time) and if during this process if a receiver observes congestion
it fall back to previous layer while exponentially incrementing the join
time. This enables receivers to cater for the heterogeneity in links and
changes in the network traffic due to congestion. In order to evaluate the
performance of RLM, author has carried out simulation using ns2 and
observing the packet loss rates.

The idea of RLM is really interesting mainly because it makes use of
existing IP Multicast Network and addresses the problem of adapting to
individual link capacity without requiring explicit support from
underlying network. Also the ideas of ‘join experiment’ and ‘shared
learning’ are quite attractive. The one problem which I foresee is the
complexity of implementing such a solution on a large scale and its
ability to handle adapt to provide fair allocation to each session in the
presence of multiple senders and receivers.
Received on Tue Oct 31 2006 - 01:14:32 EST