Analyzing the delay performance of server-based and receiver-based local recovery approaches for reliable multicast

Local recovery approaches for reliable multicast can be classified into a server-based and receiver-based approaches. This paper presents a delay analysis of server-based (L1) and receiver-based (L2) local recovery approaches for reliable multicast. We focus on the average delay between the initial arrival of a packet at the sender and its correct reception at a randomly selected receiver. Our analysis shows the following characteristics. (1) Except for extremely low loss, L2 usually becomes saturated under lower arrival rates than L1. (2) When the tail links per repair server increases, the saturated value of L1 decreases. However, L1 is still better than L2 even if the tail links per repair server comes to 200. (3) The delay performance of L1 will hardly change if we increase the number of tail links and keep k constant. The delay performance of L1 is more scalable than that of L2.