Accommodating Heterogeneity in a Multicast Session Through a Receiver-based Data Replication Scheme.
Résumé
A multicast session can involve multiple receivers with different capacities. To accommodate this heterogeneity, we propose a new replication mechanism which allows for a fine-grained multi-rate congestion control. In our scheme, some receivers (replicators) are responsible for data replication to a subset of receivers with lower capacity. A replicator, in the same way as a single-rate multicast source, adapts its rate depending on feedback it receives from the members of its associated subgroup. A simple partitioning algorithm is proposed to split a set of receivers into subgroups of similar capacities. This algorithm does not rely on a prior knowledge of the receivers' capacities and is executed on-the-fly as soon as necessary feedback are collected. To be more scalable and fairer with other sessions while improving the receivers satisfaction, we suggest to execute the partitioning algorithm at the routers. Analysis and simulations are performed in order to evaluate our approach, mainly by comparing it to traditional (source-based) replication schemes. Using ns, preliminary simulation results show the rapid convergence of the partitioning algorithm. Fairness of our scheme toward other flows is also dealt with.
Domaines
Informatique [cs]
Origine : Fichiers produits par l'(les) auteur(s)