Paper Review:† An Architecture for Wide Area Multicast Routing
Reviewer:† Kevin Hofstra
1. Is there a way to both reliably and efficiently route multicast packets across a sparsely and widely distributed area?† In what ways can we develop a architecture which is adaptive, does not require large overhead, and is protocol independent?
2. A presentation of the Protocol Independent Multicast Protocol (PIM).† Describes the architecture for efficient routing to multicast groups that span wide area and inter-domain internets.
A. Existing link-state and distance vector multicast routing schemes only have good scaling properties when multicast groups densely populate the network of interest.
B. Shortest path trees have low latency, where Core based trees have greater traffic concentrations.† PIM allows the multicast group to chosse their preference.
C. PIM differs in 2 ways from existing schemes in 2 ways:
i. Routers which have downstream members send an explicit join message to indicate the need for messages instead of vice-versa.
ii. Instead of data driven dense mode IP multicast tree construction PIM uses Rendezvous Points (RPís) to announce group need and declare new senders of a group.
D. Efficiency is measured in terms of the state, control message processing, and data packet processing required across the entire network to deliver data packets to the members of the group.† With PIM the tree is built from the leaves to the root with explicit notifications and announcements instead of assuming a root broadcast and expecting members to cancel and the entire tree to be constructed.
4. Rating-† 3†
It seems to be the first work in the area of making multicasting efficient over sparse wide area networks.† Much of the logic behind the PIM architecture seems very basic and intuitive.† I belive they have created a good base for further study, but seem to have just published the basics for new methods that are not breakthrough.† I think further works will build off this paper to make greater discoveries.
The theories presented in this paper seem fairly based and not well studied.† I am assuming that much if it could be implemented, but they have no study into its possible performance gains.† I think it would be much more convincing with experimental data.
5.† Systems analysts should recognize that some papers must be very basic and intuitive.† These papers are the first to be published in a field and will serve as the foundation for later studies.† Although they may lack experimental support they present the ideas so that they can be used in further in-depth experimental studies.