Nexa-Tick Data Launches Low-Latency Feed

Nexa Technologies will release this week a redeveloped version of its low-latency datafeed designed to meet the algorithmic trading needs of its hedge fund clients.

The new product, which will be run out of the vendor's Tick Data division, is called TickStream. TickStream is a re-engineered version of a feed that Nexa developed about six years ago for use in its direct market access trading products. Over the last year, partly in response to client demand, the vendor has invested in ways to cut latency out of the feed, officials say. By also moving it under the control of the Tick Data division, hedge fund clients can not only get historical data for creating and testing algorithmic models from Tick Data, but also a low-latency feed of market data to execute those models against.

The feed contains Level 1 and Level 2 data on all US equities, futures and options, as well as data from the London Stock Exchange and the Toronto Stock Exchange.

"The focus is on those doing statistical arbitrage and program trading," says Neal Falkenberry, senior vice president of the Tick Data division. "Everybody is moving to high-frequency trading and black box trading. The faster you can get quotes, the faster an algorithm can… send out an order. We're responding to that trend."

The changes to the feed fall into four main areas. First is connectivity: the vendor reviewed all of its exchange connections and upgraded its options connectivity from DS3 to larger OC3 lines. It also increased its server hardware, adding 19 new servers to the ticker plant that collects and monitors different feeds and then converts them into a standardized ASCII format. To monitor latency effectively, Tick Data built a server specifically to synchronize times between all servers in the data distribution process, including from those at the exchanges that collect, process and distribute quotes from the source, and the vendor's servers, which collect, normalize and redistribute the data, Falkenberry says.

Finally, Nexa began adding timestamps to quotes at each stage of the process. It does this on an external server rather than in the feed, where it only records the delta between each stamp. To add each timestamp to the feed itself "would add three milliseconds for each timestamp in the feed," Falkenberry says. While that might not sound like much, "add that three times in the process, and to us, that would be unacceptable latency," he says.

In addition, clients can place their algorithm servers directly in Nexa's co-location facilities—either in the main center in New York or the vendor's redundant site in Allen, Texas—and connect directly to Tick Data's ticker plant. That way, "the data only has to travel six feet rather than miles and through a number of switches. That greatly reduces the latency in the feed," Falkenberry says.

If clients chose to collect the feeds directly themselves, they would need redundant lines and feed handlers for each exchange. Falkenberry says firms could find themselves paying around $70,000 per month for the data and connectivity before they even budget in any development costs for keeping feed handlers up to date with the exchanges' requirements. This compares with $9,000 per month for co-locating in Nexa's facility, he says.

Falkenberry says the feed will be attractive to global investment banks and mutual funds who would want it for post-trade analysis, but the core client base will be hedge funds with between $200 million and $1 billion under management. He says this group doesn't want the costs of creating that data infrastructure for direct feeds themselves. "Tick Data has relationships with several hundred hedge funds, and we think there will be opportunities to sell into them for the real-time feed," he says.