Distributed scalability tuning for evolutionary sharding optimization with Random-equivalent security in permissionless Blockchain

Abstract

In spite of multiple advantages of the adoption of blockchain (BC), it still faces some integration challenges in modern applications, such as the Internet of Things. These challenges include low throughput rates in permissionless settings. To solve this challenge, several state-of-the-art works proposed sharding (i.e., partitioning) the BC infrastructure. Sharding the network into smaller shards improves the total system throughput, regardless of the node-shard assignment criteria. Most previous work applied a Random Sharding (RS) approach, i.e., randomly allocating nodes into predefined shards, to satisfy the required unpredictability property of node-shard allocation. In this paper, we propose a Blockchain Optimized and Secure Sharding (BOSS) protocol that aims to optimize the node-shard allocation resulting in increased throughput using a variant of the evolutionary Genetic Algorithm. The RS-equivalent levels of security and unpredictability are guaranteed by deploying a distributed random tuning mechanism for the intra-shard weight. We designed BOSS as an extension of the well-defined RS-based RapidChain protocol. We show that the proposed methods can be adapted to other sharding protocols that originally used RS techniques. We implemented and tested our protocol with more than 362,880 cases that covered seven configurable system and optimization parameters. Our evaluation revealed ≈17% average enhancement in scalability, along with a negligible <0.5% mean absolute difference in security levels. To the best of our knowledge, this is the first work that optimizes inter- and intra-shard scalability, with publicly verifiable solutions in permissionless BCs, while maintaining RS-equivalent security and unpredictability.