Performance modeling of stochastic networks with network coding

With network coding, a network node can code information flows before forwarding them. While it has been theoretically proved that network coding can achieve maximum network throughput, the theoretical results usually do not consider the burstiness of data traffic, delays, and the stochastic nature in information processing and transmission. There is currently no theory to systematically model and evaluate the performance of network coding, especially when nodes' capacity (i.e., coding and transmission) becomes stochastic. Without such a theory, the performance of network coding under various practical system settings is far from clear. To fill the vacancy, we develop an analytical approach by extending the stochastic network calculus theory to tackle the special difficulties in the evaluation of network coding. We prove the new properties of stochastic network calculus for network coding, which are used to derive the stochastic end-to-end delay and throughput.