From: Masaharu Kobashi (mkbsh@cs.washington.edu)
Date: Wed Nov 10 2004 - 00:05:11 PST
1. Main result of the paper
The paper proposes PIM, a wide-area multicast mechanism, that overcomes
shortcomings of previous multicast architectures and provides
scalability,
unicast protocol independence and great adaptability to underlying
network conditions.
2. Strengths in this paper
One of the greatest properties of PIM is the unicast protocol
independence.
Because of this property, it can be deployed without requiring changes to
the fundamental behavior of routers, hence easily deployable now and
adaptable to unexpected environmental changes in the future.
Another key element is it can reduce the total state in routers to be
on the order of the number of groups rather than the number of senders
times the number of groups.
It is also flexible enough to be able to switch from the shared-tree to
SPT when it is necessary to facilitate more efficient routing.
3. Limitations and suggested improvements
Fault-tolerant mechanism in shared-tree is not incorporated or at least
is not explained in the paper. Once a RP router is disabled, it seems
there is no way for another RP substitutes it. It is a critical point
and has to be incorporated in the architecture.
Decision on the threshold for switching from shared-tree to SPT (source-
specific shortest-path-tree) is hard to make in terms of both the
amount of traffic and the duration of high load. The paper does not
explain
well about this critical point.
There is no mechanism, or at least it is not explained in the paper,
to switch back to shared-tree from SPT. Without this mechanism, there
is possibility that with sporadic bursts from many sources the whole
tree becomes virtually SPT and it loses the merit of saving states.
4. Relevance today and future
The design seems to have great value in the current as well as future
Internet to facilitate multicast in an efficient way. Its "unicast
protocol independence" makes it possible to deployable even in the
changing future environment.
This archive was generated by hypermail 2.1.6 : Wed Nov 10 2004 - 00:05:11 PST