Incrementally deploying the Bloom filter-based scalable multicast protocol

Abstract

Scalable multicast protocol design based on the Bloom filter has received much attention from the research community recently. With the Bloom filter-based design, routing information is carried by the multicast data packet in the format of a Bloom filter. As the network node has no need maintaining a forwarding state for each group it is supporting, the Bloom filter-based multicast protocols are more scalable compared with the traditional IP multicast. However, as these protocols require networking nodes involved to perform special Bloom filter-based processing and the legacy routers can not be all upgraded over night, the incremental deployment solution for the Bloom filter-based multicast protocols is still an open technical issue. This paper presents such a solution for the Bloom filter-based multicast protocol that the new protocol is able to work with even only a small fraction of upgraded routers in the network, where the correctness will not be affected but some efficiency may be influenced. The basic idea of the solution is to consider the unicast path connecting two upgraded routers as a logical interface, and the corresponding packet processing and forwarding state maintenance are based on the logical interface. The topology information of the upgraded routers are not required, since the multicast joining messages will automatically set up the multicasting tree connecting upgraded routers. We show that the incremental solution will not affect favored scalability property of the Bloom filter-based multicast protocol, and can work with the fast joining scheme. Simulation results are given to evaluate the influence of the incremental deployment solution on the scalability, bandwidth efficiency and joining delay.