Fast failover of multicast sessions in software-defined networks [article]

Jorik Oostenbrink, Niels L. M. van Adrichem, Fernando A. Kuipers
2017 arXiv   pre-print
With the rapid growth of services that stream to groups of users comes an increased importance of and demand for reliable multicast. In this paper, we turn to software-defined networking and develop a novel general-purpose multi-failure protection algorithm to provide quick failure recovery, via Fast Failover (FF) groups, for dynamic multicast groups. This extends previous research, which either could not realize fast failover, worked only for single link failures, or was only applicable to
more » ... ic multicast groups. However, while FF is know to be fast, it requires pre-installing back-up rules. These additional memory requirements, which in a multicast setting are even more pronounced than for unicast, are often mentioned as a big disadvantage of using FF. We develop an OpenFlow application for resilient multicast, with which we study FF resource usage, in an attempt to better understand the trade-off between recovery time and resource usage. Our tests on a realistic network suggest that using FF groups can reduce the recovery time of the network significantly compared to other methods, especially when the latency between the controller and the switches is relatively large.
arXiv:1701.08182v1 fatcat:c4g6hmxyefe2zetgeaslttabjy