Network Coding (NC) can improve the information transmission efficiency and throughput of data networks. Random Linear Network Coding (RLNC) is a special form of NC scheme that is easy to be implemented. However, quantifying the performance gain of RLNC over conventional Store and Forward (S/F)-based routing system, especially for wireless network, remains an important open issue. To solve this problem, in this paper, based on abstract layer network architecture, we build a dynamic random network model with Poisson distribution describing the nodes joining the network randomly for tree-based single-source multicast in MANET. We then examine its performance by applying conventional Store and Forward with FEC (S/F-FEC) and RLNC methods respectively, and derive the analytical function expressions of average packet loss rate, successful decoding ratio and throughput with respect to the link failure probability. An experiment shows that these expressions have relatively high precision in describing the performance of RLNC. It can be used to design the practical network coding algorithm for multi-hop multicast with tree-based topology in MANET and provide a research tool for the performance analysis of RLNC.

In this paper, we derive a lower bound on the minimum decoding delay for convolutional network codes, which provides us with a guide line in the performance of decoding delay for convolutional network code decoders. The lower bound can be achievable by the sequential decoder introduced by E. Erez and F. Feder. Then we discuss the relationship between the network topology and the minimum decoding delay. Finally, we illustrate our results by an example.