Max Bowie: Latency Rules ... for Now

Low-latency technology in financial markets is often compared to Formula One race cars, which use the latest technologies to compete to build the fastest car. Similarly, trading is a competition: Being fastest gets you the best price that will make the most money, while being able to trade in and out of positions without delay ensures you can maximize your return.

The most latency-sensitive firms have done what they can in terms of implementing direct datafeeds via low-latency feed handlers and ticker plants, eliminating internal bottlenecks by replacing or recoding slower systems, and by selecting suppliers of networks and co-location services based on detailed performance metrics—and holding the suppliers to those metrics via service-level agreements.

Ultimate Source
Exchanges and other trading venues—the ultimate source of and destination for market data and trading activities—have been focused on reducing the latency with which their data reaches the clients. Part of this is driven by the need to offer as efficient a trading environment as possible, and to keep up with the speed at which clients now operate. But in today’s highly fragmented equity markets—where traders can choose where they execute based on which venue can get them the best price faster than another, or where they can potentially arbitrage a security traded on multiple venues based on delays in price distribution—speed has also become a differentiator for exchanges.

This is most true among European multilateral trading facilities or US-based ECNs and exchanges, but also for others who recognize the potential of high-frequency trading and geographic proximity to create fragmentation between different venues. For example, latency-monitoring vendor Correlix has struck deals with Nasdaq, Direct Edge and NYSE—which also utilizes Corvil for its LatencyStats.com portal—among others. Bats Global Markets recently expanded an internal monitoring tool, and most recently, the Singapore Exchange began using Corvil to measure latency of its new Reach trading engine and the co-location facility where it will reside.

Key to many of these initiatives is not just achieving the lowest latency, but also generating reports to clients that demonstrate latency is within agreed limits, and any capacity issues that might cause latency spikes, so that clients can adapt their trading strategies to changing market conditions. This is one of the reasons SGX CIO Bob Caisley gave for adopting Corvil, along with increased granularity, and SGX’s ambitions to be at the forefront of technology innovation—which, if a merger with the Australian Securities Exchange goes ahead, will be an important factor in the combined group’s ability to attract business from across Asia-Pacific.

But even F1 cars depend on more than speed alone—tires, aerofoils, race strategy and a speedy pit crew all contribute to the best overall performance, which might change mid-race depending on rivals’ strategies. And how long speed alone will remain a competitive differentiator is under scrutiny—if only because of the cost required to continue delivering incremental improvements.

But once you’ve accelerated data delivery as much as possible, what’s next? You could accelerate other aspects of the data flow beyond just data from liquid marketplaces, such as complex calculations and analytics—for example, by using a GPU-powered appliance to run pricing and risk calculations in a fraction of the time, as high-performance analytics provider FuzzyLogix is doing with its Tanay appliances.

Or, will latency-sensitive firms capitalize on their acceleration work by tying up exclusive deals with the connectivity providers offering the best latency, resulting in those firms “owning” trading on specific markets—at least until someone else does the same thing with a better service? The cost would be huge, but so could the gains.

So, will firms be content with having the same latency as their rivals and differentiating themselves on doing smarter things? Or—just as some F1 teams make do without a major engine supply—will we see low-latency market access move from a democratized model to one of “haves” and “have-nots”?