Aggregated Multicast Home Page

Network Research Lab

Computer Science Department

University of California, Los Angeles

Overview

Though IP multicast is resource efficient in delivering data to a group of members simultaneously, it suffers from scalability problem with the number of concurrently active multicast. We propose a scheme, called aggregated multicast, to reduce multicast states. Aggregated multicast is an inter-group tree sharing scheme, in which multiple groups are forced to share a single delivery tree. At the expense of some bandwidth wastage, this scheme can reduce multicast state and tree management overhead at transit routers. It also can simplify and facilitate the provisioning of QoS guarantee for multicast in future aggregated-flow-based QoS networks.

Motivation

IP multicast utilizes a tree delivery structure, on which data packets are duplicated only at fork nodes and are forwarded only once over each link. This approach makes IP multicast resource-efficient in delivering data to a group of members simultaneously and scalable in supporting very large multicast groups. However, a multicast distribution tree requires all tree nodes to maintain per-group (or even per-group/source) forwarding state, and the number of forwarding state entries grows with the number of "passing-by" groups. As multicast gains widespread use and the number of concurrently active groups grows, more and more forwarding state entries will be needed, especially in transit domains. More forwarding entries translate into more memory requirements, and may also lead to slower forwarding process. Thus, though IP multicast scales well to the number of members within a single multicast group, it suffers from scalability problems when the number of simultaneous active multicast groups is very large.

Our Approach

Recently, the state scalability problem has prompted some research proposals, either eliminating multicast states at (some) routers or aggregating multicast states purely in routers. We propose a novel scheme, called aggregated multicast, to reduce multicast states. Unlike previous approaches, our scheme forces aggregated multicast multiple groups to share one distribution tree. In our scheme, core routers need to keep states only per aggregated tree instead of per group. This can significantly reduce the total number of trees in the network and thus reduce forwarding states.

The benefits of aggregated multicast scheme:

While this approach significantly reduces the number of forwarding state entries and alleviates overhead associated with tree management, it may also waste bandwidth as it delivers data to non-group-members, that is, if “leaky match” is allowed. There is thus a trade-off between control overhead savings via aggregation and bandwidth wastage introduced by common tree sharing. To find a best compromise point, we use a group-tree matching algorithm.

Protocols Designed

We have designed three protocols based on aggregated multicast approach

As to details of the protocols, please refer to our papers.

More will come about the design and implementation of the protocols.

Back to top

Journal/Magazine Papers

Conference/Workshop Papers

Technical Reports and Internet Drafts

Posters

Back to top

Related Links, Projects and Papers

Back to top

Questions and Comments to: jcui AT cs.ucla.edu      Last Modified: January 13rd, 2003