Determining Optimal Shard Size in a Hierarchical Blockchain Architecture

Abstract

Monolithic blockchain architectures employed in Bitcoin and other major alt-coins are inherently non-scalable. In recent past, some hierarchical approaches have been explored to shard the decentralized blockchain to improve scalability. However, there is no discussion in the literature about how to determine an optimal shard size to maximize performance and how the presence of malicious or faulty nodes can impact on choosing an optimal shard size. To address these issues, this paper presents a sharding scheme and validation protocols for a hierarchical blockchain architecture named OptiShard. The hierarchy divides the network nodes into multiple disjoint shards and the majority of transactions are distributed among these shards in non-overlapped fashion. Optimal shard size is determined based on two parameters: performance and correctness of transaction validation in the presence of malicious or faulty nodes. OptiShard provides guaranteed majority of good shards, subject to a maximum allowable threshold of faulty nodes, by choosing the right shard size. It also provides a mechanism for identifying faulty shards and discarding all their transactions through the overlapping of a small fraction of transactions across all the shards. Experimental results performed on up to 800 Amazon EC2 nodes conform to the theoretical performance analyses and exhibit the impact of sharding the network on performance.