STRESS: efficient multicast shared trees via restricted search

A shared tree with a lower end-to-end delay carries out multicast packet delivery more efficiently. This article introduces efficient multicast shared trees via restricted search (STRESS) realized by means of locating the closest on-tree nodes for new group members to join the trees. Such a tree ensures an end-to-end delay shorter than that in a tree built by connecting new members always to a fixed tree node (like CBT or PIM-SM). STRESS is shaped only by multicast group members to yield a tree as small as possible, totally avoiding the difficult task of determining a fixed tree node for member joining. It requires no centralized point to keep track of all on-tree nodes, therefore creating no performance or reliability bottleneck. Two mechanisms for restricted search over multicast trees plus local search are considered and evaluated by simulation. STRESS is shown to be more efficient than CBT, as a result of connecting new members to their nearest known on-tree nodes.