Maximizing Performance of Blockchain Systems via Sharding and Collective Signing

Abstract

In this thesis we present a novel blockchain consensus protocol based on sharding. Sharding enables blockchain systems to be scalable in terms of nodes and performance, but adds many additional problems. We tackle many of those issues to present a system that is capable of achieving the throughput requirements of the increasingly popular distributed applications, and decentralized currencies. First, we explain the needed background and related research needed to understand the work presented. Then we introduce a novel consensus protocol based on our previous work, Concordia, integrating it to build a new system capable of high transaction throughput and low confirmation latency. Using the proposed Concordia as the main building block, we then create a management layer that supplies all the requirements needed for its operation in a sharding context. We then solve the key issues of this sort of system such as cross-shard transactions, randomness generation, and block propagation, showing original methods and state-of-the art technology. Our performance study validates the proposed system and shows its capabilities. Analyzing the results, we confirm that the proposed protocol enables hundreds of nodes to participate in the system, and is capable of achieving scalable throughput with low confirmation latency.