From: Jonas Lindberg (jonaslin@kth.se)
Date: Tue Nov 09 2004 - 12:18:38 PST
Review of S. Deering, D. L. Estrin, D. Farinacci, V. Jacobson, C-G. Liu, and
L Wei's "The PIM Architecture for Wide-Area Multicast Routing"
By: Jonas Lindberg
In this paper the authors describe a new architecture for multicasting
traffic. The authors begin with describing weaknesses in earlier multicast
protocols. Link-state protocols scales badly as a consequence of the
requirement that every router need to have membership information for every
group in its domain. Distance-vector protocols is inefficient in the
"sparse" case (where the network is sparsely populated by receivers) since
it send the traffic to all routers unless these routers sends prune messages
with a given interval. The core based tree distributes all its traffic over
one tree, which can result in congestion problems when multiple sources send
traffic over the same path at the same time.
The solution presented in the paper consists of combining several techniques
to a protocol that requires receivers to explicitly join groups and supports
multicast over both shortest-path trees and shared trees. The authors
provide good reasoning about why PIM is better than the other protocols in
different situations, but I do not think they really motivate their design
choices. More test results that show how PIM actually performs would have
been interesting.
This archive was generated by hypermail 2.1.6 : Tue Nov 09 2004 - 21:19:23 PST