A Scalable Destination-Oriented MulticastProtocol with Incremental Deployability

Abstract

In this paper, we develop a scalable destination-oriented multicast (DOM) protocol for computer networks where the routers have enhanced intelligence to process packets. The basic idea of DOM is that each multicast data packet carries explicit destinations information, instead of an implicit group address, to facilitate the data delivery. Based on such destinations information, each router can compute necessary multicast copies and next-hop interfaces. A fundamental issue in DOM is to constrain the bandwidth overhead due to explicit addressing, which is tackled with a Bloom-filter based design. Our design incorporates the reverse path forwarding (RPF) concept and the BGP routing information, so that DOM can work efficiently in practical networking scenarios especially with asymmetric inter-domain routing. A critical issue in Bloom-filter based design is the issue of forwarding loop due to false positives. We propose an accurate tree branch pruning scheme, which equips the DOM the capability to completely and efficiently remove the false-positive forwarding loop. Furthermore, we study how the DOM can be deployed in an incremental manner over a network, in which only a small fraction of the routers have DOM-aware intelligence while others are legacy routers. We present extensive simulation results over a practical topology to demonstrate the performance of DOM, with comparison to the traditional IP multicast and the free riding multicast (FRM) protocols.