Following a period of subdued activity post-pandemic, the IPO markets are regaining momentum. Looking back at Alibaba’s record launch in 2014, valued at over $169 billion, we now find the company once again in the spotlight with the press reflecting on its prospects of taking two of its subsidiaries to IPO in 2024 and subsequent abrupt share price decline. IPOs are not straightforward. But the most pertinent risk for this piece is how IPO trading offers a gruelling test of Capital Markets’ participants’ IT infrastructure.
The importance of Performance Analysis
CEO of Capital Markets infrastructure specialist Beeks Group, Gordon McArthur clearly states: “IPO trading can be so busy that it can seriously expose weaknesses in a company’s data handling performance. Real-time data is crucial – and when real time is measured in nanosecond units you need to be able to capture, analyse and replay data with extremely high precision to be able to fully understand what’s happening when your system is misbehaving.”
Beeks prides itself on being Capital Markets’ dedicated Infrastructure-as-a-Service (IaaS) provider since 2011. Over the years they have honed the technology to support, monitor and protect trade executions and trading performance metrics in Capital Markets, Banks, and Financial Services companies.
For this, the organisation provides the tools and services to ensure businesses have a high level of timely visibility of where in their trading network problems are occurring. They also have the expertise and knowledge to resolve performance issues in an agile manner, thus mitigating loss of profitability.
“It’s not always easy to see what’s happening under the trading bonnet, and it’s dangerous to take a bargain basement approach to performance monitoring, especially during high-profile IPOs,” McArthur asserts. “A lot of businesses run the risk of being duped by their data. While things might look like they’re working properly, without the proper infrastructure it isn’t possible to see that things have broken until it’s far too late, and a lot of money has been lost.”
A cautionary tale
To illustrate this, Beeks Analytics Chief Technology Officer (CTO), Steve Rodgers recalls a particular issue at a market participant as they tried to engage in the trading around Alibaba’s debut IPO back in 2014. He says: “The institution lost a significant amount of money during the Alibaba IPO and couldn’t work out how. With no real-time monitoring and a reliance on reactive replays from their application logs, they had no visibility of what had actually gone wrong. They couldn’t see any faults, the market data ticker plant reported processing 100% of the data with no loss downstream to trading applications. Yet they still made a significant loss during trading.”
Outlining how Beeks got involved, Rodgers says: “At the time the institution was aware of Beeks as a vendor for their performance lab and so they invited us to troubleshoot. We found that their incumbent hardware-based market data ticker plant was to blame.The Alibaba IPO was causing microbursts the ticker plant could not keep up with when operating in low latency mode. Live market data feeds were overwhelming the system and causing delays. All the data was eventually delivered, but delayed just enough to give the bank a false position of what was happening at those extremely busy periods of market activity.”
Beeks identified that the initial attempts to reproduce the problem from feed handler logs were unsuccessful because the microbursts were not captured correctly at the time and ended up being flattened out when replayed. It took a high-fidelity historical capture of the original trading day, replayed with nanosecond precision to reproduce the fault reliably.
The birth of the mdPlay
In low-latency environments, performance analysis and network monitoring solutions are vital components. High-fidelity packet capture and precision replay capabilities give firms greater scope for more targeted drilling down and quicker resolution when issues arise.
“In 2014 we used mdPlay to demonstrate the original fault to their ticker plant vendor who were eventually able to understand and resolve the problem,” says Rodgers. “This experience effectively gave rise to our unique mdPlay managed service offering. At the time mdPlay represented an all-in-one system to carry out all its own capture, replay and scenario management delivered as a turnkey managed service,” he says. “In the intervening decade, mdPlay has been automated, giving end-users instant visibility of scenarios on-demand and the ability to manage the tool themselves.”
mdPlay capabilities vs competitors
Traditional solutions would read the recorded market data directly from storage and in doing so introduce latency and jitter which was not present in the original. These solutions also smooth out peaks in market data which would have potentially been the cause of application issues.
mdPlay takes a different approach. It spools up the replay and pre-caches everything, resulting in maximum fidelity and accuracy in reproduction. The replayed scenarios are published to the network exactly as they were originally transmitted during live trading: a true like-for-like reproduction. mdPlay also replays data through the same infrastructure and code as the production path, which helps rule out or identify infrastructure causes.
mdPlay also allows a library of scenarios to be built up to automate testing of system behaviour against known problematic market conditions – essential for QA and backtesting of algorithmic trading systems. In addition, mdPlay can alter the replay speed with precision – answering “what if?” questions such as ‘what if the markets were twice as fast’, or ‘what if we move from 10Gbit to 40Gbit networks?’
“As with all trade executions, Capital Markets businesses need a robust strategy, plan and infrastructure capability to take advantage of the best prices, whether public launch events or daily tick to trade,” comments CEO McArthur. “It’s a delicate and ongoing balance between timely trade and vigilant monitoring. mdPlay forms part of our suite of Analytics tools that enable exactly that.” In conclusion, McArthur says: “Essentially our competitors can point out problems. With mdPlay, you can anticipate and prevent them.”