Time for Tools: Finding and Removing the Slightest Delay
August 17, 2009
In an industry where a few extra milliseconds of latency can cost millions, the key to survival and ensuring customer satisfaction is technology. Technology that lets trading firms know and understand exactly where the most miniscule delay might be, so it can be addressed and removed.
The rise of high-frequency trading and speed-of-light transactions has fueled the need. Now, 70 percent of average daily trading volume in equities involves a buy or sell order from a high-frequency trader, according to research Aite Group (see Data Sweep, p. 23).
"Only in the last 12 months are we really at a widespread level getting to a point where latencies are so low it takes pretty sophisticated technology to even measure it," said Kevin McPartland, senior analyst with Tabb Group. "If you are talking about microsecond differences and trying to understand where connections are fast and slow, it gets down to pretty technical issues, and how to be able to do that accurately."
Firms are on a spending spree for technology that identifies latencies. Last year, total spending software and services for measuring delays inside and outside a firm's own data centers came to $241 million, according to the Tabb Group. Spending on these tools is expected to grow from $277 million in 2009 to $390 million in 2011.
"Despite the fact that people have been trying to reduce latency for a number of years, we still have a long way to go," said McPartland. "It's still a key concern for the majority of trading firms, whether their strategies are ultra-low latency or not. That being the case, now that we're getting into a level of microseconds, it's even harder to see that your improvements are paying off. It's requiring much more sophisticated technology to track microsecond differences."
The increasing need for better monitoring and measurement began with the recent emphasis on high frequency trading, said Donal Byrne, CEO of Corvil, a provider of latency monitoring and managing software who's clients include the CME Group and Fixnetix, a provider of ultra-low latency market data.
"The nature of algorithmic and high-frequency trading is that it's quantitative in nature and it's based on saying if I do A., I'll get B., and any time you have lack of predictability in that then the outcome of your algorithm is also going to be uncertain."
With tools that determine the extent of a latency issue between two points in the trading infrastructure, firms can start to tune their trading strategies based on these latency numbers, explained Adam Honoré, research director with Aite Group. "Part one of this is reducing your own latency, part two is gaming the latency that you anticipate others having and part three of this, if you are a liquidity provider, especially on the dark pool side, is that you might want to provide latency statistics to your users," he said.
It's part two-measuring and managing the external latency-that Corvil competitor Correlix said its RaceTeam service concentrates on. RaceTeam focuses on the external environment, showing customers what they cannot see today and incorporating more parties into a team that can work collaboratively in solving latency issues, said Shawn Melamed, CEO of latency measurement software provider Correlix.
