High-Performance Automated Trading Network Architectures
Firms in the automated trading business recognize that having a low-latency infrastructure can improve the profits of the business. In this paper, we explore the key considerations to implementing an ultra- low-latency network and highlight some key misconceptions when evaluating the suitability of network platforms for automated trading businesses.
Trading in financial markets is increasingly being performed by software programs. These programs run a variety of businesses including proprietary trading for a firm and services such as best execution for the firm’s customers. In particular, the term “high-frequency trading,” or HFT, has gained popularity to describe the most intensive of the automated trading businesses.
A competitive advantage in all these businesses is to be able to access the liquidity pool faster than the competition. To this end, firms are focused on lowering the latency of the end-to-end trade flow.
From an exchange perspective, there is desire to provide deep liquidity pools with the quickest access to the market. With securities now trading on multiple venues, being the liquidity choice is instrumental to the exchange’s competitiveness.
While the absolute latency is often a focus, the capability of the infrastructure to deliver predictable latency and handle ever-increasing volumes is also a critical requirement. Market data volumes continue to increase at exponential rates. Business factors such as smaller order sizes and reduced tick sizes are driving the increases in data volumes.
Firms will generally disaggregate the trade flow into functional components—application, middleware, server, network—and each functional owner will attempt to remove latency from their area, while improving the predictability and increasing the message capacity of their portion of the trade flow.
Finally, financial firms see the ongoing enhancements to the infrastructure, especially ongoing latency reduction, as a requirement to sustain a competitive advantage.