But while colocation facilities are a hot topic in high-frequency trading circles, they address only one piece of the latency equation: the distance between the server and the exchange, points out George Hessler, EVP of Lime Brokerage, an agency broker that offers a centralized data center with high-speed fiber access to multiple venues. For example, "It doesn't really take into account how fast the provider feeds market data into the client's server, or how fast the market data is converted from the different exchange protocols and filtered so that it's useful for the black box to process," he says. There is also the issues of how fast a broker-dealer and the exchange can process the orders generated by the black boxes.
According to Hessler, one of Lime's strengths is that it normalizes all of the data feeds that use various exchange protocols via Citrius, Lime's low-latency market-data feed. "Therefore each client only has to write to our Citrius API and they can take market data," he explains.
Further, instead of processing a fire hose of data containing every symbol, Lime only sends updates on the symbols that pertain to the firm's strategies, Hessler says. "We end up cutting down on the amount of data," he asserts. "Otherwise the firm is looking at huge amounts of data, and it slows down the processing of its systems."
Accelerating Executions With CEP
As the demand grows for faster executions, firms are turning to complex event processing, or CEP, technology to detect patterns in real-time data. Typically the CEP engine sits on the black box server in the colocation facility.
"With-high frequency trading you're analyzing flows of market data coming in against complex patterns that indicate trading opportunities, and you're placing orders in the market in real time," explains John Bates, founder and general manager at the Apama division of Progress Software, whose CEP engine is used to build algorithms that detect patterns in streaming data.
"You need to be looking across every market and every dark pool, and the only way to do that fast enough is to take in direct feeds from every venue," notes Kevin McPartland, senior analyst at TABB Group. He says firms need tools such as CEP to enable a direct link between their algos and the data feeds "so that the execution can be very, very fast."
Of course, managing all of this rapid-fire activity creates its own challenges. Traditionally buy- and sell-side firms have relied on order management systems (OMSs) and execution management systems (EMSs) to direct and monitor trading activity. But, experts say, automated trading systems need faster platforms, such as FTEN's high-frequency execution platform, which the company claims can execute between 2,500 and 8,000 orders per second.
"If you're sending 1,000 orders per second, you can't watch that on a blotter," says TABB's McPartland. "In fact, automated trading is too fast for the human eye to monitor." Recently, he notes, Nasdaq upgraded the INET matching engine to execute orders in 250 milliseconds.
"There ought to be more creative ways to see what you're doing," adds McPartland, who predicts that visualization technology will come into play. "If you are trading the S&P 500 and every bar is tall or short or green or red," this could be a way to monitor high-frequency executions, he suggests.
Ultimately, the push toward zero latency will drive further advances in high-frequency trading technology, and firms seeking to be first will continue to invest in the latest low-latency solutions. But while faster and cheaper technology has made high-frequency trading more accessible to smaller firms, it's by no means a cheap proposition, points out McPartland. "I think you could do it for a few hundred thousand dollars," he estimates. But as firms scale out to the multiple product types and geographies, he notes, they need more data center space, networking equipment and databases, and the cost escalates.
Switch and Data's Panzica refers to a high-frequency trading firm that is overhauling its infrastructure. He says the firm, which is moving from two data centers to four, is spending $5 million for the refresh.
