Linear Scalability from Sharding and PoS

Abstract

Scalability is one of the most important problems in blockchain and has been the focus of both industry practitioners and academic researchers since Bitcoin was born. The blockchain has insufficient ability in handling large-scale concurrent transactions. The more transactions that are processed in the network, the more scalability problems appear in the network. Compared to the transaction throughput achieved by the electronic payment channels of mature development, the limitation is magnified. In this paper, we propose a novel consensus mechanism from sharding and Proof-of-Stake (PoS)—a scalable blockchain model that supports high concurrency, and achieve the linear expansion of transaction processing scale while ensuring security in order to prove the feasibility of the proposal. The proposal views the blocks in the network as two levels, namely, intermediate transition blocks (i.e., middle blocks) and final confirmation blocks (i.e., final blocks), and takes epoch as the basic unit of the consensus mechanism operation cycle. An epoch is a recursive process for each cycle of the consensus mechanism operation. Each epoch is equipped with four types of interactions, namely, node sharding, transaction sharding, internal consensus, and final block generation, and thereby determines the network status via main chain and shard chains. Given n nodes in the network, we construct one Validity group and p- 1 Regular groups (each one contains n/p nodes). The regular groups create transition blocks according to the transaction pool; the validity group extracts information from the transition blocks created by the regular groups, generates and sends final confirmation blocks to the main chain. PoS consensus is exploited to ensure that adversaries are not able to launch attacks on specific shards (neither transaction nor node shards). We also describe how to re-group the nodes in and add new node to the network. We provide the security analysis under several standard attack models.