Performance of random network coding for data dissemination

Abstract

Network coding is an alternative to traditional store-and-forward routing and is known to be necessary to achieve network capacity. It has also been shown randomized network coding is robust, and far outperforms store-and-forward for multicast. While much focus has been on the data rates achievable with coding, we focus on the time needed to broadcast a finite amount of data throughout networks using distributed randomized linear coding. We consider networks with increasingly complex graphs. We use analysis of the dissemination time using coding in the line network to discuss the performance of coding in networks with more complex topologies, such as the Manhattan grid network..