Patterned Erasure Correcting Codes for Low Storage-Overhead Blockchain Systems

Abstract

Traditionally in blockchain systems, the sequence of transactions acts as a decentralized ledger where each node in the network is required to store an identical copy. As a result, the storage space required to store the entire blockchain can become prohibitively large. Erasure coding and network coding techniques were previously introduced to mitigate this storage burden. In this work, we first observe that node failures in the network can occur periodically. Next, we take advantage of this property to develop a coding scheme which corrects a predefined set of possible node erasure patterns. By explicitly focusing on certain erasure patterns, proposed codes have lower storage requirements than erasure codes which correct all possible erasure patterns, and moreover, our codes achieve the minimum average storage per node under the given assumptions.