Bandwidth-Efficient Delay-And Loss-Tolerant Overlay Routing

Peer-To-Peer (P2P) systems are mostly deployed in heterogeneous environments with resource availability varying not only across the nodes, but also over time. If any of the shared computational, storage or network resources are exhausted, failures and delays occur. The commonly used crash-stop failure model assumes that once a node stops sending messages it never again resumes. Such failures are trivially detected and appropriate algorithms are run that maintain the connectivity and routing efficiency of the P2P overlay under continuous arrivals and departures of the peers (i.e. churn). The failure detection mechanisms in the crash-stop model are typically tuned to minimize the number of false positives that might be caused by intermittent message dropping or delays.