Collaborative Management of Correlated Multicast Transfer

Abstract

Multicast is designed to jointly deliver the same content from a single source to a set of destinations; hence, it can efficiently save the bandwidth consumption and reduce the load on the source. Distributed file systems in data centers provide multiple replicas for each data block. In this case, the traditional Multicast faces the diversity of senders since it is sufficient for each receiver to get a replica from any sender. This brings new opportunities and challenges to reduce the bandwidth consumption of a multicast transfer. This chapter focuses on such Multicast with uncertain senders and constructs an efficient routing forest with the minimum cost (MCF). MCF spans each destination by one and only one source, while minimizing the total cost (i.e. the weight sum of all links in one multicast routing) for delivering the same content from the source side to all destinations. Prior approaches for deterministic Multicast do not exploit the opportunities of a collection of sources; hence, they remain inapplicable to the MCF problem. Actually, the MCF problem for a multi-source Multicast is proved to be NP-hard. Therefore, we propose two $(2+\varepsilon )$-approximation methods, named P-MCF and E-MCF. We conduct experiments on our SDN testbed together with large-scale simulations under the random SDN network, regular SDN network and scale-free SDN network. All manifest that our MCF approach always occupies fewer network links and incurs less network cost for any uncertain Multicast than the traditional Steiner minimum tree (SMT) of any related deterministic Multicast, irrespective of the used network topology and the setting of Multicast transfers.