Latency standards for exchange and arbitrage trading are now measured in microseconds. Non-functional testing for the stress points is a critical challenge. Exactpro, owned by the London Stock Exchange (LSEG), has quickly become established as an industry leader.
Beyond the LSEG, Exactpro’s clients include Tradition, the leading inter-dealer broker and derivatives specialist. What these clients share in common is the challenge of connecting trading systems in highly complex markets. For example: how can a trading platform test early on in the software development process for small defects caused by minute fluctuations in price that have not been correctly captured in a post-trade data feed?
According to Iosif Itkin, CEO and co-founder at Exactpro, the solution to testing the resiliency of those systems, and finding any potential faults, is to replicate the chaos of markets. “We try to look at the system from the eyes of an algorithmic trader instead of a network that we are trying to break. It’s a slight change of perspective — the confluence of functional and non-functional testing.”
One key tool used by Exactpro to test trading systems is its team of so-called ‘mini-robots’. “The idea is to use a simplified version of an algo trading engine to simulate market conditions,” said Itkin. “The mini-robots are not exact replicas of algos, but simplified models used to create realistic trading scenarios easily and quickly,“ he added.
One client that employs this technique is Tradition. “The most important aspect of non-functional testing is accurately mimicking real-world conditions so that we can prepare for what might actually happen,” said Jeremy Norwood, project and programme change manager at Tradition. “This is daunting for a brokerage firm because we operate in an extremely complex financial ecosystem. If we imagine banks as complex systems in their own right, then a brokerage is where these systems intersect, and they interact in chaotic, hard-to-predict ways.”
According to Itkin, Exactpro is now looking to leverage machine learning techniques to monitor the output of a system in order to predict where defects are likely to be found. “We want to design a system that can find errors without being told. It can use historical data to narrow-in on family resemblances in unusual patterns of data, distinguishing between natural fluctuations and defects,” he said. “This is a more fruitful approach than just trying to crash a trading platform which operates at the scale of the LSEG — which is extremely robust anyway.”
Exactpro is still a young company; launched in the US in September 2009 and in Russia in December 2009 by Iosif Itkin and Alexey Zverev and a small team of specialists. “The team had a vision of what we needed to do in order to successfully test the complex technology systems and platforms underpinning the financial markets,” said Itkin. The team developed a diverse set of tools for automated testing, including a multi-protocol functional test automation tool, a simple but extremely powerful load injector, and also a tool for conducting passive testing to help with exchange client on-boarding and certification.
The London Stock Exchange was one of Exactpro’s first customers, and went on to acquire the business at the end of May 2015. Exactpro now employs 550 staff, and Iosif Itkin is confident that its customer market will continue to grow. “We believe that appropriate investment in risk controls and rigorous testing are critical to ensure stable and efficient functioning of the financial markets,” he said. “A higher percentage of the overall budget allocated to security and compliance will be the ongoing trend.”