Summary: An improved throughput-competitive algorithm for Online
Dept. of Computer Science
Johns Hopkins University
We present an O(log 2 n log M) throughout competitive algorithm for online multicast, where n is
the number of nodes in the graph, and M is the number of multicast groups. This improves upon the
previous best O((log n + log log M)(log n + log M) log n) competitive algorithm by Goel, Henzinger
and Plotkin. and closes the gap between the upper and lower bounds to O(log n).
This algorithm uses the recent extension of the maximal dense tree algorithm to graphs with
dynamic edge weights.
Our algorithm is modular, and does not involve an integrated analysis of its winner picking and
1.1 Problem denition
Multicasting in a communications network is the broadcasting of a signal from its source to a subset of
nodes in the network. Point to point connections, or unicast, are a special case of multicast.