Overlay multicast (Application-Level Multicast (ALM)) constructs a multicast delivery tree among end hosts. Unlike traditional IP multicast where the internal tree nodes are dedicated routers which are relatively stable and do not leave the multicast
tree voluntarily, the non-leaf nodes in the overlay tree are free end hosts which can join/leave the overlay at will, or even crash without notification. So, the leaving node can leave suddenly and cannot give its descendants (and the Rendez-vous Point (RP)) the time to prepare the recovering (the reconnection) of the overlay tree, and so there is a need to trigger a rearrangement process in which each one of its descendants should rejoin the overlay tree. In this case, all of its downstream nodes are partitioned from the overlay tree and cannot get the multicast data any more. These dynamic characteristics cause the instability of the overlay tree, which can significantly impact the user.
A key challenge in constructing an efficient and resilient ALM protocol is to provide fast data recovery when overlay node failures partition the data delivery paths. In this paper, we analyze the performance of the ALM tree recovery solutions using different metrics.
