We present a fault tolerant scheme for recovery from single or multiple node failures in multi-directional multicast trees. The scheme is based on cyclic structures providing alternative paths to eliminate faulty nodes and reroute the traffic. Our scheme is independent of message source and direction in the tree, provides a basis for on-the-fly repair and can be used as a platform for various strategies for reconnecting tree partitions. It only requires an underlying infrastructure to provide a reliable routing service. Although it is described in the context of a message multicast, the scheme can be used universally in all systems using tree-based overlay networks for communication among components.
Aiguo FeiJun‐Hong CuiMário GerlaDirceu Cavendish
Paulo CoelhoTarcisio CeolinAlysson BessaniFernando Luís DottiFernando Pedone