Data Centres in HFT & the Impact on the SOR Landscape

The role of data centres in High Frequency Trading and its impact on the SOR landscape

Historically the world of High Frequency Trading (HFT) has been shrouded in mystery, with few people outside the industry understanding what they do or how they operate.

Written by Darren Watkins, Managing Director Published Tuesday, 18 August 2015 08:00

Historically the world of High Frequency Trading (HFT) has been shrouded in mystery, with few people outside the industry understanding what they do or how they operate. In fact, most HFT companies operate in complete secrecy with names unfamiliar to anyone except the employees themselves or the exchanges on which they ply their trade.

That is of course, until Michael Lewis' book, Flash Boys, lifted the lid on what businesses operating at this end of the equities markets have been doing.

In laymen's terms High Frequency Trading is a type of algorithmic trading characterized by high turnover and high order-to-trade ratios. Specifically, it is the use of sophisticated technological tools and computer algorithms to rapidly trade securities.

Although HFT has traditionally been very popular – accounting for over 70% of trade volumes in 2010 – it has since dropped to below 50 per cent after 2012, indicating a shift in the financial trading landscape.

So what's caused this, and more importantly, why should this impact data centre strategies?

The answer is simple. As a result of the financial crisis, large institutional investors have decided to err on the side of caution, relying less on HFT techniques, in favour of longer term trading strategies driven by Smart Order Routing (SOR) principles. The reason being – they are seen as more stable, and thus a safer bet.

As such, while HFT focus on the second by second trades, and are reliant on technology to trade huge volumes and move in and out of trading positions in fractions of a second, SOR does not. Unlike HFT, SOR algorithms don't rely on the micro-second activities of the HFT world and instead use programmatic strategies to take longer term positions in the securities markets, which are also more widely used in multi-asset classes.

As a result, rather than the close proximity to trading exchanges essential to HFT, SOR techniques rely instead on deterministic latency connections, which are more cost effective than those used by HFT, and don't need to be so close to the exchange. That being the case, many cost conscious firms would be well placed to put their SOR engines at the virtual 'centre-of-gravity' of the asset class being traded.

With this in mind and given the distributed nature of the asset classes across the London metro footprint from Slough, through the City and the Docklands, a data centre located within the M25, can just as efficiently offer the low latency connections required to access all the necessary exchanges in a colocation environment. Furthermore, it can do it at a fraction of the cost of the 'finance eco-systems' which data centre operators promote as a premium, but costly financial services location.

It's true that HFT strategies located at exchange locations offer the ultimate performance at a premium – but why operate all of your strategies and back office functions from a premium priced location when not all of them require it? It seems like overkill and a waste of money. In fact, not only can significant cost savings be made by adopting a distributed IT deployment but it also comes with the additional benefit of increased resilience as a result of using multiple sites.

With the rapid changes in technologies at every level of the IT stack there is value in challenging the traditional 'all your eggs in one basket' deployment models of the past. Many of these deployments were implemented because of the belief it was simpler to keep everything together – a sales and marketing message supported by some data centre operators to their own benefit.

But, don't believe all of the hype and be clear on why applications sit in a particular location. If they don't all need to be there, is there a better, more efficient and more cost effective solution out there? In 2015, the answer is always yes.