文摘
Nowadays, several advanced ICT infrastructures demand multicast services able to jointly provide a high degree of reliability and good performances despite of failure occurrences. The current practice to tolerate failures in such services mainly consists in the adoption of methods based on temporal redundancy, where messages are successfully delivered to all the interested destinations by means of retransmissions. However, in such methods, a high degree of reliability is typically obtained at the expenses of severe performance fluctuations and instability. On the contrary, approaches that make use of forward error correction imply a more performance-friendly behavior; however, the effective tuning of such approaches is a critical and challenging task, which has limited their applicability in large-scale multicast services. This paper is aimed at presenting a solution to resolve this issue by effectively tuning the redundancy degree to be applied, so as to be able to tolerate the losses affecting the multicast communications without an excessive, and un-needed, workload imposed over the network. To this aim, we drawn from the experience of layered multicast, commonly applied to multimedia content delivery, for addressing such tunability problem. Specifically, we detail how to apply layered multicast within the context of a well known multicast service and how to determine the correct redundancy degree to be applied. We also make use of extensive simulations to evaluate the quality of the proposed approach in terms of reliability, performance and overhead, and to compare it with the best retransmission-based approach available in the literature.
