Black box traders want-and need-greater bandwidth to execute their high-volume and quant strategies. Traditional voice and Internet protocol telecommunication lines just can’t handle the sheer volume they churn into the marketplace. To execute their statistical arbitrage and pairs strategies, these traders need information highways with unlimited space and autobahn-style speed limits if market data, trades, and even clearing and settlement are going to move at lightning speed. As a result, these information aqueducts-or ultra low-latency communications networks-are red hot.
A select group of traders is betting on them, and the stakes are large. High-volume proprietary trading through algorithms and black boxes demands that latency drop into the realm of milliseconds. High-volume traders fear that without these fast networks, they will miss out on fills.
But help is on the way. Some vendors are looking to fill this niche and hoping to garner the business from these black box and quantitative traders.
Brennan Carley, chief strategy officer and global head of product operations for NYFIX, a New York-based provider of front-to-back office trading technology, explains the reason behind the demand for ultra low-latency networks: It’s just not the sheer number of trades that require the expanded bandwidth, Carley says. The demand is also driven by the high number of cancellations.
“As markets become both more electronic and more fragmented, many traders place orders in multiple exchanges or ECNs, and then need to cancel (or change the size) of those orders when they get filled (or partially filled) somewhere else,” Carley says.
Traffic Jam
These so-called cancellation rates are creating more data traffic-hence, the greater demand for ultra-low latency. Looking forward, the onset of Regulation NMS in the U.S. and Markets in Financial Instruments Directive (MiFID) in Europe may slow down routing and increase latency even further, all because more data have to pass through the pipes.
The move to quoting options in pennies in the U.S. is also going to have an effect on how many messages are being delivered at any one point in time. Larry Tabb, founder and CEO of the consultancy TABB Group, says the introduction of penny options at the end of last month will begin to push quote traffic up from a current peak of around 190,000 messages per second to more than 500,000 messages per second in the options world.
“This is a huge issue,” Tabb says, and although a high-speed network is part of the answer, he believes many high-volume trading firms will still need to co-locate their black boxes.
Co-location is a strategy in which a trading firm places its black box or algorithmic trading solution and servers in an exchange or ECN’s data center (or very close to it). This reduces the latency associated with long-distance telecommunications transmission to nearly zero.
New York to Chicago
Ultra low-latency networks are also designed to diminish transmission delays, but are limited by necessity to one geographical area (the typical latency between Chicago and New York is 35 milliseconds each way.).
Tom Chippas, head of Deutsche Bank’s electronic equity trading platform, dubbed Autobahn Equity, says there is a need for this kind of solution because firms under Reg NMS won’t be able to put a box at one market center and solve all of their latency problems. “Historically co-location was the key to reducing latency between the customer and the market center,” says Chippas, who’s in charge of the product in North America. “Common thinking today is that you will have to have low-latency access to all of them.”
To address this, financial services connectivity provider BT Radianz is building a super-large pipe’ connectivity network in the New York City metropolitan area. Due to be launched this month, it will offer a super-fast, managed Ethernet local area network that runs directly from large banks and hedge funds to the exchanges where they trade.
This solution, combined with BT Radianz’ existing Proximity managed hosting and co-location product, supports those of its customers’ trading strategies that require extremely low latency, says Mark Akass, the firm’s CTO.
This low-latency BT Radianz network will take care of speed issues customers have at proprietary and statistical arbitrage trading desks that transact huge volumes and deliver even greater message traffic.
Case by Case
Chippas notes that a very low latency New York metro-area network is a viable solution, especially given that most equities venues are located in the area. Other asset classes will also require low-latency networks in the near future, says Akass, and other metropolitan areas may be looking for the same low-latency treatment. BT Radianz will consider installing the network in new areas on a case-by-case basis, he says.
There are a number of solutions to address high-speed connectivity. These include direct leased-line “dark fiber” or Ethernet connections; extranets, or shared networks; and virtual private networks. New York-based Transaction Network Services (TNS) supplies data communications solutions for transaction-oriented applications. It uses an encrypted network (which uses Internet protocols) built purely for electronic trading.
Alan Schwartz, executive vice president and general manager of the financial services division at TNS, calls it “the first many-to-many network.”
TNS has two offerings that are similar to the new BT Radianz product, Schwartz says: its low latency Managed Direct Connect solution and its Fiber Solution. The Managed Direct Connect product is available over a choice of high-speed infrastructures- including dark fiber and local area network solutions like Ethernet.
Dark fiber refers to unused strands of fiber within a fiber-optic cable. Firms can lease dark fiber from telecommunications companies and “light” it by using it to communicate directly to chosen venues such as exchanges.
Some large hedge funds have even been discussing leasing their own dark fiber for connectivity to the exchanges, Tabb notes.
But firms doing this find that the dark-fiber route is costly. It requires a skilled technical staff to put the architecture into place and to monitor the management tools that connect applications to the network.
The Cost
James Parker, vice president of sales engineering and product development for the financial services division at TNS, says, “Firms need to focus on their core competency. You have to look at the cost and reality of dark fiber. The majority look at it and ask themselves, Is it worth spending this much?’ If they are in the metro area they can get fast technology without that.”
NYFIX’s Carley says traditional IP networks provide a good solution for a large portion of the marketplace. It is mainly proprietary trading desks at brokerage firms and hedge funds that are looking for speed to handle their high-velocity trading, mostly for statistical arbitrage and pairs trading. That’s because they are looking to take advantage of market inefficiencies that only last for mere milliseconds. “If it lasts 20 milliseconds, and I can do it in two milliseconds, great. If I can only do it in 30 milliseconds, it is too late,” he says.
Carley doesn’t believe many trading firms could justify buying and lighting their own fiber. That’s because the cost is prohibitive, and besides, the exchanges don’t want lots of fiber coming from a multitude of firms. “They are saying, Come through an extranet.’ “
Carley is also a board member of California-based Yipes, a global provider of managed Ethernet solutions, which he says has done good business providing Ethernet connections from customers to the exchanges.
TNS’s Parker says there’s interest from the sellside to “kick it up a notch” for their largest buyside customers and deploy the high-speed solution to a few of them. However, he points out, “The infrastructure has to be there; otherwise, the cost is too high.”
Parker says the market for very high-speed, ultra low-latency routing is tiny. “The difference in speed is only significant for a very small number,” he says. “Even then, some of them say it is not worth it. It is hard to believe they can make enough to cover the costs.”
Chippas agrees; “It is not for all players. For arb strategies, yes. For less time-sensitive trading strategies, no.”
