Coordinated cooperative task computing using crash-prone processors with unreliable multicast.

This paper presents a new message-passing algorithm, called Do-UM, for distributed cooperative task computing in synchronous settings where processors may crash, and where any multicasts (or broadcasts) performed by crashing processors are unreliable. We specify the algorithm, prove its correctness and analyse its complexity. We show that its worst case available processor steps is S=Θt+nlognloglogn+f(n−f) and that the number of messages sent is less than n2t+nf2, where n is the number of processors, t is the number of tasks to be executed and f is the number of failures. To assess the performance of the algorithm in practical scenarios, we perform an experimental evaluation on a planetary-scale distributed platform. This also allows us to compare our algorithm with the currently best algorithm that is, however, explicitly designed to use reliable multicast; the results suggest that our algorithm does not lose much efficiency in order to cope with unreliable multicast.