To cope with the electronic trading explosion, firms must achieve a balance between direct raw data feeds at high speed from exchanges, and indirect, clean data that arrives milliseconds later.

Now more than ever, speed is key: the explosion of electronic trading, the exponential growth of data, the proliferation of hedge funds and regulatory stress are increasing the pressure to reduce latency. Traditionally it has been the brokers who have had to worry about, and invest in, speedier trading technologies. As the buy side takes increasing control of its own execution, however, latency has become a pressing concern for investment managers as well. With diminishing portfolio yields and the growth of hedge funds operating high speed automated strategies, being able to seize the opportunity is a crucial competitive advantage. While firms scramble for margins of milliseconds, however, the challenges of achieving low latency are physical, technical and infrastructural, and will generate losers as well as winners.

Mike Powell, global head of realtime enterprise information, Reuters, comments: "The assumption is that the faster you process the models and execute the orders, the better the results you get in the market. This is partly a factor of how quickly the customers, applications work, but it is also a matter of how quickly the feed handlers and APIs can process the data from an exchange."

As Philip Slavin, head of European product strategy for primary trading platform Fidessa, explains, processing the data from the exchange is one of the more challenging parts of the overall process. "Where market data latency is concerned, the key thing is the number of 'network hops'; however fast one hop may be, the overall process will always be constrained by the slowest hop. Black boxes require the input of market data, upon which trading decisions are made, and instructions generated. In support of this process an infrastructure is required, naturally creating additional hops which "automatically makes things slower", Slavin explains. The recent introduction of FAST FIX suggests, for example, that people have realized that market data is going to be one of the slowest hops in this process and needs a low latency solution."

Identifying where exactly the weak link resides is also problematic for those firms deploying multiple solutions from multiple vendors, says Simon Garland, CTO of data capture software provider Kx Systems, because "each one will say their application is the fastest. If there are any performance problems the vendors will always point the finger to some other part of the solution, especially if you have three or four conflicting vendor solutions fighting for scarce time and space resources on the same machine."

More complex automated trading, in particular the growing movement toward cross-asset and multi-asset trading, will only compound the latency issue. Firms looking at different asset classes or cross-asset trading in a particular geography or group of exchanges are not necessarily co-located at or near the exchanges, explains Kevin Covington, head of product development at BT Radianz. "This creates the problem of how you achieve low latency trading on multiple exchanges efficiently; you would have to repeat the process, systems, et cetera to each one, and they may not be near to each other or on the same development lifecycle." For firms running algorithms that depend upon low latency connections to exchanges, this is only just the beginning of understanding the complexity, Covington warns: "At some point the limitations of physics become involved."

In response to these problems, firms are already moving away from their data aggregators, choosing instead to take in raw data sourced directly from the exchange, or even "locating applications at the exchange thereby minimizing the distance the information has to travel", says Mark Hepsworth, head of Comstock, Interactive Data. For data feed handlers this is good news, as John Coulter, vice president of marketing and business at Vhayu Technologies, provider of streaming tick processing software, is happy to admit. "Taking the data direct from the source is going to become the predominant way of receiving data, not through a data aggregator, and as a result vendors that sell data feed handlers and tick processing are going to rise in number." He adds: "A lot of firms are now taking their data in directly from the exchanges, eliminating the vendors, to benefit in terms of latency and to secure lower execution fees."

Publicly at least, the data aggregators are not convinced that the equation is this simple, however. Interactive Data's Hepsworth comments: "There are those for whom low latency is a factor, but not in an extreme way; other factors, such as a wide range of sources and communication costs are more important. Many people want a whole variety of different sources of data; they may only take a bit of data from each source so that they have a wide range of content."

Taking in raw data directly from the exchange means sacrificing quality for speed, while implementing this direct linkage has its own costs, and can prove intricate and laborious, as Covington points out: "Each exchange is slightly different; each configuration for each exchange is slightly different, and each way of communicating with each exchange is slightly different; so there's a lot of work involved in getting those solutions done." Furthermore, developing infrastructure outside the firm, near the exchanges, means managing more of the firm's resources in remote locations, which is complicated and risky.

Despite the immediate difficulties of direct feeds, more and more firms will be forced to make the investment. Where data aggregators cannot compete on speed therefore, they will have to market themselves on quality, as Garland suggests: "Data vendors now have to offer a completely different added value because there are faster ways to get data directly from an exchange." Consolidated data feeds (whereby many global exchanges and content sources are blended into a single feed using a common data model) are one such value added offering to which the vendors are turning. Powell explains: "Consolidated data feeds provide significant economies of scale, access to OTC data and news, plus the benefits of a consistent symbology and data model all via a single feed."

Choosing between direct raw feeds at high speed, and indirect, clean data milliseconds slower, with added analytics and news to boot, is a tricky balancing act. Garland notes that many firms are ultimately opting for both, in order to provide a check on their raw data. "We have some customers who will have the direct feed to an exchange but they'll have Reuters or Bloomberg too, so it's a back-up but not the primary data source - because they want more speed." He continues: "Although Reuters and other data vendors are more reliable, firms will take the chance of going with the faster feed, even though it's more expensive."

Powell confirms that many Reuters customers are taking a more structured approach towards their latency data needs: "Many are using consolidated feeds with intelligent traffic management for desktops and applications which are not latency sensitive, and installing direct feed connections for high end trading applications. lt's not a case of one or the other. It's an issue of fit for purpose and managing total cost of ownership."

For firms as well, the picture looks complicated and costly as the demands of low latency put their market data infrastructure under stress. Higher update rates and faster processing require higher spec networks and hardware to support it, says Powell. "lt's not just a case of whether the vendors can provide low latency; it's also a question of how firms build out their networks and manage their costs in the face of increasing demand for speed and depth of data."

The big firms are already making significant investments in terms of their staff and skills, Covington observes. "Certainly the bigger guys are probably realizing that if they are going to be successful, there is a huge amount of investment that's going to be required." He adds: "If you don't have the investment or wherewithal to do that, you might be left out in the cold. We might see people being forced to re-think how and whom they do business with," he concludes.

©2006 STP Information Services. All rights reserved. Used by permission.