An investigation of simulating blockchain networks : an abstract approach

Abstract

This project aims to build upon existing research into the modelling and simulation of blockchain networks, focusing specifically on abstracting properties of the network into a conceptually and computationally simpler model. The model proposed by this model was created using the python-based library PyCATSHOO, allowing for the simulation of the entire network on a single machine. The system is constructed using deterministic finite state machines, with all network communications being simulated through delays. Monte Carlo simulations were used to test different configurations of the model together with performance indicators which continuously analyse the state of the simulator and return a value. Focus was placed on recreating and extending the features of the state of the art whist improving the efficiency of the model. IT constructed from the ground up to follow a high level approach, abstracting away low-level details of the network which should not affect the accuracy of the simulator. The model created could reproduce existing simulation data as well as data observed for existing deployments without physically implementing a network or creating a complex virtual network on a machine. The simulator allows for various properties and indicators to be observed, such as block sizes, block intervals and propagation times, and blockchain splits which result orphaned blocks. The evaluation focuses on the scalability of the model, observing the effect of a growing network size on the performance and security of the network. The resulting data highlighted how blockchains such as Bitcoin are capable of scaling up to thousands of nodes and still maintain strong consensus and resistance to attacks. Finally, improvements were proposed to the implementation to improve its efficiency and broaden the scope of the properties it may simulate.