From: Charles Reis (creis@cs.washington.edu)
Date: Tue Nov 09 2004 - 23:59:58 PST
The PIM Architecture for Wide-Area Multicast Routing
Deering, Estrin, Farinacci, Jacobson, Liu, Wei, 1996.
In stark contrast to the previous two data-heavy measurement papers we have read, this paper focuses almost entirely on proposing a new architecture. The authors address multicast routing in a sparse, wide-area network, since they show previous multicast techniques to be dependent on a "dense mode" assumption which scales poorly. Their proposal improves scaling for sparse groups by having recipients notify routers of groups of interest (maintained with soft state), rather than having routers forward by default and prune out useless paths.
This new PIM protocol is also designed to take advantage of existing approaches when appropriate, switching between shortest path trees (SPT) or shared trees depending on the characteristics of the traffic, as well as interoperating with other multicast protocols and working independent of unicast routing choices.
While the system is described in considerable detail, no evaluation is given to show whether the design meets their stated goals. The authors claim a prototype has been built, but no quantitative results are shown, nor are particularly strong analytical arguments made about why the proposed mechanisms achieve their goals. Without further validation, it seems that the paper merely presents a point in the multicast design space, with a good formulation of the issues to consider in such a system.
This archive was generated by hypermail 2.1.6 : Wed Nov 10 2004 - 00:00:12 PST