From: Shobhit Raj Mathur (shobhit@cs.washington.edu)
Date: Tue Nov 09 2004 - 22:10:54 PST
The PIM Architecture for Wide-Area Multicast Routing
====================================================
Protocol Independent Multicast (PIM) is a protocol for multicast which is
independent of any particular routing protocol. Prior multicast routing
protocols were intended for use in 'dense' networks where a group is
widely represented. These protocols suffered from scaling problems in
environments where a relatively small fraction of routers wanted to
receive traffic for a particular multicast group. This paper describes a
protocol for efficient routing in such 'sparse' networks.
Compared to the total number of destinations in the Internet, the number
of destinations belonging to one group is very small. A protocol to
address this issue is required as the previous protocols are poorly suited
to such a scenario. To address this, PIM assigns a rendezvous point (RP)
to each group. All the routers in a domain know the unicast IP address of
the RP for a given group. A multicast forwarding tree is built as a result
of routers sending 'join' messages to the RP. PIM is capable of
constructing two type of trees - Shared tree and source-specific shortest
path tree (SPT). The normal mode of operation creates the shared tree
first, followed by one or more source-specific trees if there is enough
traffic. I will skip the details of the protocol and discuss the merits.
PIM is 'protocol independent' since all of its mechanisms for building and
maintaining trees are independent of the routing protocol used. Other
protocols were derived either from link-state or distance-vector routing.
PIM uses receiver initiated membership advertisement and hence is scalable
in sparse networks. It reduces the total state in routers from (number of
senders X number of groups) to an order of groups. It supports both shared
and SPT trees. While shared trees help in scalability, SPT trees move
towards optimal and efficient routing.
The paper does not describe clearly what is the criteria for moving from
shared to SPT trees. It also does not contain any simulation results which
would help the reader compare the performance with earlier protocols. The
paper gives a good motivation for the need for a protocol like PIM. Though
the protocol seems reasonable, a more theoretical and experimental
approach would give the paper more credibility.
This archive was generated by hypermail 2.1.6 : Tue Nov 09 2004 - 22:10:55 PST