A Cloud-Native Distributed Real-Time Multi-Agent Stock Exchange Simulator for Investigating Inter-Market Arbitrage

We describe a new multi-agent distributed stock exchange simulation environment (DSXE), built to model the global network of contemporary financial markets. DSXE is an advance on existing state-of-the-art simulation platforms available in the public domain: it is a modular and highly configurable environment which allows researchers to setup and deploy stock exchange agents and trader agents in different geographical locations using commercial cloud-computing services. The efficient implementation in C++ enables the runningof large-scale simulations with many simultaneous traders. DSXE has been successfully used to model fragmented markets and to demonstrate price convergence resulting from arbitrageurs operating between two exchange venues trading the same asset. We report here on a series of experiments conducted to measure and quantify the performance and scalability of the system. The implementation successfully achieves the goal of modelling high-frequency trading, demonstrating the capability to process up to 355 messages per second. Finally, potential avenues for further research and suggested improvements to the implementation are outlined. The C++ code for DSXE is being made available as open-source code via GitHub.