Abstract :
We propose a network information flow strategy of optimal multicast with erasure correction coding (OMEC), in which erasure correction is a means of utilizing diversities rather than combating packet losses. An information source is first encoded by an erasure correction code and then optimally multicast by a server to sinks. The OMEC scheme has the following characteristics: (1) It is at least as good as optimal routing and, for some cases, can realize a network throughput arbitrarily larger than that achievable by pure routing. (2) For most scenarios, the approach yields a throughput comparable with the optimal throughput achievable with network coding. In fact, for most examples in the literature for which network coding makes high throughput gains compared to pure routing, OMEC performs close to the optimal. (3) Unlike general network coding, intermediate nodes are only required to route information. This makes OMEC compatible with currently deployed network structures. (4) Surprisingly, unlike pure optimal routing, OMEC can be efficiently computed in polynomial time. While OMEC is, in general, suboptimal compared to full fledged network coding, it enjoys most of the computational and throughput benefits associated with network coding over pure routing, without requiring any changes to existing hardware infrastructures or protocol stacks.