There is an increasing demand for performance and scalability in micro-transactions between people and in the machine-to-machine environment. The number and frequency of transactions are likely to only increase, as will the use of decentralized ledgers and blockchain technologies to address this.

However, there are challenges to obtain sufficient performance and scalability in ledger-based consensus systems. There are proposals for off-chain, layer-two, solutions that may provide some mitigation of the load. However, there is still a need to improve the performance of the main blockchain.

This paper focuses on a proposal known as ScaleGraph, where data sharding is explicitly built into the system. This sharding is then used to form validator groups, which are able to perform parallel transactions without the need for post-shard synchronization.

The contribution lies in simulating and analyzing the ScaleGraph proposal to test the behavior of ScaleGraph and highlighting potential issues. The simulation is written in Go with a locally simulated network in order to simplify the simulation and remove network overhead. However, while testing the artifact, it was found that some key assumptions made about finding close nodes in the underlying DHT did not hold for the implementation. These assumptions and their impact on the system is discussed and some potential methods of mitigating them are brought forward.