A Proposal for Blockchain Account Network Partitioning Optimization

Blockchain technology, originally developed to support cryptocurrencies, faces a barrier to wider adoption because of its limited processing throughput. Sharding is garnering attention as one of the solutions to address this problem. There are two primary issues associated with implementing sharding in blockchain. The first issue is the high ratio of cross-shard transactions (TXs) to all TXs, and the second issue is an imbalance in TX processing volumes across the shards. These issues are in a trade-off relationship that could significantly reduce blockchain throughput. In this paper, we propose an account network partitioning algorithm based on soft clustering using Stochastic Gradient Descent (SGD) to achieve more optimal sharding. The proposed algorithm is implemented, and the results of evaluation experiments using real transaction data are described.