Latency special report

Click here to download the PDF

Speed Is Nothing Without Control

Since trading firms first realized there was an advantage to be gained by being able to capture market data and trade faster than their rivals, an arms race has been underway to see who could engineer the fastest and most efficient trading infrastructure with as little latency as possible.

First, firms turned to direct datafeeds, captured using low-latency ticker plants, encouraging exchanges to distribute their feeds even faster. Then firms implemented low-latency messaging layers to distribute that fast data throughout their organizations, which in turn placed pressure on vendors to streamline their technology and eliminate even the slightest trace of latency. This turned the industry's attention to hardware-based technologies that function faster than software, and eliminating network layers altogether by moving trading servers into proximity hosting datacenters close to exchanges, and ultimately into the same facilities as exchange matching engines, spawning a wave of megadatecenters, such as those being built by NYSE Euronext, CME Group and the Singapore Exchange.

But as low latency becomes ever-more important, it also becomes harder to achieve. Saving milliseconds a few years ago was easier than saving microseconds or nanoseconds today, and firms are using sophisticated timestamping and measurement tools to monitor latency at the most granular level possible-in some cases, allowing them to generate microsecond improvements overall by eking out nanoseconds here and there. In addition, firms looking for end-to-end views of latency are also demanding transparency into the delays introduced by links in the chain outside of their control, such as trading venues themselves, prompting exchanges to roll out their own latency monitoring solutions, to provide members with statistics such as order-to-trade and tradeto- tape latency, and where a firm's overall roundtrip latency stands in relation to its peers.

Simply put, those who can't measure their performance can't perform, and will have to look elsewhere for trading advantage. But how long can it be before latency is exhausted as a competitive differentiator? Ultimately, this arms race is limited by light speed and the laws of physics. And perhaps in turn, those firms that can't compete on latency will have the upper hand when the rest of the industry gives up battling over picoseconds and femtoseconds. But according to the participants in this report, there's still a long way to go before we need to worry about that.

Click here to download the PDF