Stretching BFT

State-of-the-art BFT protocols remain far from the maximum theoretical throughput. Based on exhaustive evaluation and monitoring of existing BFT protocols, we highlight few impediments to their scaling. These include the use of IP multicast, the presence of bottlenecks due to asymmetric replica processing, and an unbalanced network bandwidth utilization. To better evaluate the actual impact of these scalability impediments, we devised Ring, a new BFT protocol, which circumvents them. As its name suggests, Ring uses a ring communication topology, where, in the fault-free case, each replica only performs point-to-point communications with two other replicas, namely its neighbors on the ring. Moreover, all replicas equally accept requests from clients and perform symmetric processing. Our experiments show that on a Fast Ethernet network Ring achieves an aggregate throughput of 118 Mbps, which is 27% higher than most efficient state-of-the-art BFT protocols. Ring approaches but does not reach the throughput theoretical maximum. Yet, its very performance makes it possible to envision a new generation of BFT protocols that might reach the actual theoretical maximum. Keywords-BFT protocol, fault-tolerance, throughput