Abstract
Wireless mesh networks offer a cost-effective multi-hop backhaul, but their lossy links degrade multicast performance. We refine the M/G/1 delay model of two active reliable multicast protocols, AMRHy (hybrid) and DyRAM (receiver-initiated): every node serves data packets, ACKs and NACKs; lost control packets are retransmitted hop by hop; NACKs are repeated with an exponential back-off; and NACKs triggered by an upstream loss reach the tail relay in bursts. The model states its validity domain in closed form. A packet-level discrete-event simulation of the same mechanisms, with idealised replier election and ACK suppression, validates the model within 8% inside this domain (27 configurations per protocol). At the reference rate of 0.05 packets/ms, AMRHy repairs a lost packet in 19.5 ms against 44.9 ms for DyRAM, whose receivers detect a loss only when the next packet arrives. The delay ratio falls from 7.4 at 0.01 packets/ms to 1.6 at 0.1 packets/ms. With independent losses, losing control packets as often as data packets increases the delay by at most 10%, and NACK bursts limit both protocols beyond 52 to 53 receivers per group; bursty losses increase both delays and widen the advantage of AMRHy. These results support hybrid loss recovery, especially for low-rate sessions.