Reach High Performance Consensus on a Multicasting Network with Dual Failure Mode
王淑卿, 嚴國慶, 鄭建富, Shu-Ching Wang, Kuo-Qin Yan, C.F. Cheng · 2004
The distributed consensus problem is one of the most important problems in designing a fault-tolerant distributed system. In prior work, the consensus problem was solved in a Fully Connected Network (FCN) or a Generalize Connectivity Network (GCN) with fallible communication media and cost two rounds of message exchange to solve the consensus problem. However the above network topology is not practical in the real world. So we revisit the consensus problem in a MultiCasting network (MCN) which is the most practical network topology in the real world, and we also propose two new protocols Relay Fault-Tolerance Channel (RFC) and Efficient MultiCasting Agreement Protocol (EMAP). Using RFC, we only need one round of message exchange to get enough information in FCN, BCN, GCN and MCN. We also enlarge the fault tolerant capability by allowing both dormant faults and malicious faults exist in a multicasting network. In this paper, the protocols that we proposed use only one round to exchange messages and can tolerate the maximum number of allowable faulty communication media.