You’re probably already familiar with the concept of high-frequency trading (HFT). And if not, the basic ideas is that traders use hardware and software to:
- Receive signals – in as close to real time as possible – that indicate the most advantageous time to buy and sell securities.
- Initiate those transactions.
Millions of dollars stand to be won by the traders who are quickest to the draw.
On any given day, investors may move a billion or more shares using HFT. And behind every single transaction is a labyrinth of information technology and networking infrastructure. From servers, to coding languages, to big-data processing platforms, to fiber-optic cables (and, increasingly, microwaves and lasers that literally beam data between stock exchanges) – all of the core technology fueling a trader’s chances of profiting converge in one location: the data center.
Ever since the late 2000s, when HFT entered the mainstream, colocation data centers have been the engines for traders – and exchanges – all over the world. The reliability, cost-efficiency and performance of these facilities is crucial in determining investors’ outcomes.
To give you a better sense of just how important colocation data centers are to modern trading, let’s look at some of the key technological drivers.
Low latency is king
High-frequency trading is all about location, location, location. Why? Because of latency. A delay of just a few milliseconds is enough to miss out on millions of dollars. Light can only travel so quickly through fiber-optic cables. You can’t do very much to make it go faster, but you can position your HFT technology stacks as close to the exchange’s stacks as possible. This reduces the distance light has to travel over data lines between the exchange and your computing resources.
The location of those computing stacks is therefore of the essence. This is why, for instance, many investors find Sabey’s SDC Manhattan facility so appealing. It’s located just a few miles from the New York Stock Exchange’s servers.
For additional context on how important latency is to traders, consider this: A blink of the eye lasts roughly 400 milliseconds. And yet, according to The Quarterly Journal of Economics, a mere 3 milliseconds is an “eternity” in the world of HFT. In the time it takes you to blink an eye, countless millions of dollars have been won and lost.
A peak at the stacks
The equipment inside data centers is just as important as the infrastructure between them. First and foremost, data centers must connect to high-speed networks – because how else would data signals and subsequent transactions travel between the exchanges and your HFT technology stack?
Zooming in, the day-to-day performance of the facility is just as essential. Mission-critical servers must not fail while the exchanges are open. As we mentioned earlier, a lot happens in just a few milliseconds. Imagine if there was power or cooling failure, or other facility maintenance issue that resulted in server downtime for just a few minutes. Considering the stakes, the track record of a colocation facility is of the utmost importance when it comes to selecting a home for HFT stacks.
Density is also a factor that needs to be considered. Artificial intelligence, machine learning and other data analytics workloads put a lot of strain on facility infrastructure, particularly on cooling. Not all facilities can cost-effectively support the high-density GPU clusters needed to sustain such awesome processing power, and that can be problematic – especially considering the technology arms race underway in HFT. High-density, power-hungry racks will only get denser and hungrier as large investment firms try to one-up each other.
Therefore, traders need to work with colocation data centers that can support today’s stacks while also being able adjust to future loads. Because tomorrow always comes a little sooner than expected.
Just the tip of the iceberg
In the coming years, we fully expect to see greater progression in HFT technologies. Microwave technology, for example, started shaving milliseconds off round-trip transmission time as early as 2011, because light travels faster through air than through glass. Then came the lasers, circa 2014. They’re a little more weather-resistant than microwaves.
What comes next? That’s not really for us to say. The only thing we know for certain is that, with each new advancement in the chase for zero-latency transmissions, there will always be one constant in the world of HFT, and that’s the data center.
