" class="no-js "lang="en-US"> EXCLUSIVE: ‘Smarter and faster?’ – Ashok Kalyanswamy, Saxo Bank and Milan Dvorak, Magmio in ‘The Fintech Magazine’
Thursday, March 28, 2024

EXCLUSIVE: ‘Smarter and faster?’ – Ashok Kalyanswamy, Saxo Bank and Milan Dvorak, Magmio in ‘The Fintech Magazine’

FPGAs have driven tick to trade, but is speed always the most important benchmark in investing? Ashok Kalyanswamy of online trading and investment specialist Saxo Bank, and Milan Dvorak, Founder of Netcop Technologies, now Magmio, agree to differ Ashok Kalyanswamy, Saxo Bank | Fintech Finance

They’re used by physicists at CERN to pick up instabilities in the Large Hadron Collider particle beam; they could help a fibre-optic camera process multispectral images of arteries fast enough to be useful to heart surgeons; and they can eliminate the infinitesimal time delay experienced by scientists trying to study the beat of a fly’s wing.

‘They’ are field programmable gate arrays (FPGAs) – logic elements on a chip that can be configured and reconfigured for tasks whose success depends on eliminating the temporal time jitters seen with conventional computing. In the financial world, they are most commonly associated with ultra-low latency investment transactions, the ‘tick to trade’ metric that determines success in high-frequency environments. But that’s not their only application.

While in prop trading, using FPGA technology undoubtedly accelerates the race to zero latency, Milan Dvorak, founder of Netcope Technologies, which now provides FPGA services to the investment industry under the Magmio brand, suggests that banks, brokers and other institutions could also find value applying them as risk gateways or market data processing gateways, especially in periods of high volatility. And everyone is now familiar with those.

“During COVID, many of these institutions ran into performance limitations other than latency, just dealing with the sheer amount of data,” he says. “And that’s another area where FPGAs can help.”

Ashok Kalyanswamy, chief information officer at online trading and investment specialist Saxo Bank, had previous experience of incorporating the hardware at a broker house that co-located FPGAs at exchanges. While hugely expensive at the time, it did put a market-beating distance between the company and its rivals because it was able to accelerate the last mile in the trading process. But he says that you have to be discerning about where you apply the technology. And, as a retail-focussed investment bank, facilitating other people’s trades, Saxo hasn’t yet seen a compelling reason for using FPGAs.

“You have to think about what you are striving for,” says Kalyanswamy. “If ultra-low latency is key to your strategy, you co-locate as close to the exchanges as possible and you can get raw market data and trade directly; if you have a proprietary algorithm and you’re trying to achieve ultra-low latency, then absolutely, [FPGAs are useful]. However, if you are catering for a retail investor who’s in Australia, trading with a broker on Nasdaq, it doesn’t really make sense just to optimise the last mile, because it’s many hops from the person making the decision to trade, and the order hitting the market.

Milan Dvořák, Netcope | Fintech Finance

If you are sitting in the UK, and you’re trying to trade on the Japanese market, is it going to make a nanosecond latency difference? Not really. By the time you hit the button on your app or desktop and it travels first to your broker and then to the market, those latencies are bigger than the last mile in the FPGA.

“When you talk about ultra-low latency trading, to optimise the hardware, co-locating the decision-making, your algorithms and your strategies are as important as the connectivity part. You have to think, end-to-end, who is making the decision to hit the button to trade.”

In Saxo Bank’s case, we are a global facilitator in trading and investment, connected to 88 markets around the globe in 25-plus countries. We always try to get the best price for the customer, react to customers’ orders and what they are trying to achieve, as quickly as possible.

”But we look at the whole value chain: from different end points around the world, the network we connect to, the latency we have in our infrastructure throughout. We have a more of a systems-approach to latency and making sure there are no inefficiencies in our pipeline to the market, as opposed to optimising the last mile.”

For the majority of Dvorak’s clients, though, that last sprint is all-important.

“We basically create a tailored solution to a very specific use case and, in trading, the use case is usually ultra-low latency,” he says. “So, there is no operating system, there is no overhead, FPGAs are just pure logic, programmed to do one thing, and that’s trading. That’s why this technology can achieve great performance and low latency, beating the standard software solution by a long shot. But, at the same time, a FPGA is still programmable, so you can change it on a daily basis; upload new strategies, adjust your logic every day. You’re not stuck with an expensive piece of hardware that will eventually be outdated or stop functioning.

“FPGAs bring the best of both worlds together,” says Dvorak. “They give you ultra-high performance with programmability and flexibility.”

In this scenario, traders have a choice – to build or buy.

“Clients usually see two main challenges,” says Dvorak. “One is time. If you are starting from scratch, it will probably take you a long time to get something up and running on FPGA hardware. The other difficulty is finding the right people to do the job, because there aren’t many hardware designers or FPGA engineers around, and, if there are, they are probably already working for your competition.”

Before tasking the internal IT team with the project, it’s worth noting that coding for FPGAs is probably going to be very unlike any other programming they’ve undertaken. FPGAs are configured using a hardware-description language, and while that’s not difficult, it requires a different mindset. While most software code is procedural, gateware is descriptive.

“So, it comes down to a classic build versus buy equation,” says Dvorak. “You can decide to build, which probably means two-plus years spent in development, with all the upfront costs, because you need to hire the people, buy the hardware. Or you can buy, really decrease your time to market and have a better cost control, but with some kind of dependency on the vendor.”

Kalyanswamy doesn’t deny the value of FPGA technology to some market participants, but he believes the future of trading is about much more than speed.

“What traders are trying to do is execute a strategy,” he says, “and they have many different parameters in which to do that. A holistic approach to supporting them includes giving them timely and relevant information. We are all in the business of information and risk: information to make sound decisions; risk management so you can really manage your portfolio. Those are the things Saxo is focussed on and they are really critical to successful trading and making money.

“Our business is to understand our traders, what their strategy is and give them the information to help them trade. That’s not just market data, but also news, it’s earnings and it could be information on Twitter. And when they want to make trading decisions, we give them sophisticated tools so they can send the orders as quickly as possible.

We also give traders risk management tools that equip them with a real-time view of what the risk is.”

“It’s a balance between smart and fast,” agrees Dvorak. “Traders need more information; they need to make educated decisions. But that also means they need better technology to handle that. FPGAs are getting more capable, so it’s fair to expect people will want to do more with them in the future. That’s where Magmio comes in – to help reduce the complexity about using and programming FPGAs.”


 

This article was published in The Fintech Magazine #22, Page 72-73

People In This Post

Companies In This Post

  1. Yapily Named a Supplier on Crown Commercial Service’s Open Banking DPS Framework Read more
  2. allpay Appointed as Official Supplier on Crown Commercial Service’s Open Banking and Fund Administration & Disbursement Services Dynamic Purchasing Systems Read more
  3. Oliver Wyman Announces Mariya Rosberg as Americas Head of Banking and Financial Services Practice Read more
  4. Alchemy Pay Invests in UK Fintech LaPay and Secures API License as Part of Global Web3 Expansion Read more
  5. QNB Introduces FAWRAN for Fast Payments Within Qatar Read more