Three market data themes have emerged at the SIFMA show this week: low cost; low latency; and the emergence of new, more diverse types of market data and data services.
The quest continues for high throughput and low latency delivery of data and news for the algo traders who use the information to fuel their automated trading strategies. "Latency is always a concern," notes Terry Roche, executive vice president, information management solutions in the market division of Thomson Reuters. Market participants — especially exchanges, for whom low-latency promises are a lure for order flow — continue to seek to drive their message rates up and latency down. Today, for instance, stock market Direct Edge announced its use of 29West's messaging platform for its exchange infrastructure. Direct Edge says its market share of U.S. equity trades has grown from just under 5% in October 2008 to over 12.5% currently, making it the third largest stock market in the country, after the New York Stock Exchange and Nasdaq. Using 29West's messaging backbone, the exchange says it has improved throughput while reducing its infrastructure footprint by half.
Today, NYSE Technologies announced its new Market Data Platform V5. This ticker plant technology, according to NYSE, offers 10 times the performance of previous generation market data platforms while allowing customers to reduce their hardware footprint by as much as 80%. NYSE and Intel, on whose Xeon 5500 chips the product runs, say the platform can process 500,000 messages per second on a single CPU core, with a mean latency of less than 40 microseconds over Remote Direct Memory Access and a range of 10-20 microseconds over Local Direct Memory Access. "Latency was the most important consideration in the development of this product," says Conor Allen, head of R&D at NYSE Technologies.
Tervela announced yesterday a messaging platform that the company says can deliver 4 million to 64 million messages per second with a consistent sub 10 microsecond platform latency, driven by Tervela's new ASICs, FPGAs and processors.
There are signs, though, that Wall Street's emphasis on throughput and latency may be shifting. At an event last night, a bulge-bracket firm IT executive noted, "There's a difference between high message rates and being able to execute. Algorithmic traders don't necessarily need 1.5 million messages a second, they might need 100,000 messages." For instance, if a trading strategy only involved IBM, Intel and Apple, there would be no need for the trading program to be fed market data for other stocks. "How to filter data is the key thing," he said.
Wall Street firms are seeking lower-priced market data products and services and taking a harder look at who in their organizations really needs access to expensive data feeds. In a June 2 webcast, Interactive Data surveyed 180 audience members from large buy-side and sell-side institutions, regional firms and hedge funds. Nearly 58 percent of respondents indicated that their biggest market data challenge was reducing overall spend. A third (32 percent) of respondents said the initiative they are getting most of their cost savings from is eliminating redundant providers. More than a third (37 percent) said market data expenses in 2009 compared to last year remained stable (-5% to +5 %). "We're seeing large institutions re-architect for total cost of ownership," for instance by taking advantage of multithreading, agrees Roche.
"Firms are trying to manage network costs," says Chuck Thompson, president of eSignal, an Interactive Data division. His company's Market Q data service provides real-time "snapshots" of market data to brokers in increments of time such as a second, rather than data streams that contain every tick. Ticker symbols brokers are uninterested in can be set to "hibernate" (not be listed) to further conserve bandwidth.
"We're seeing firms try to reduce market data costs," echoes Jeremy Sokolic, senior vice-president of sales and marketing at CarryQuote. "People are getting more price sensitive." This is a self-serving observation, since CarryQuote provides a market data service for fund managers that costs between $20 and $100 a month, cost-wise a middle ground between Thomson Reuters' and Bloomberg's subscription-based market data feeds and free data feeds such as BATS' U.S. equities feed and delayed stock quotes. "There have been Rolls Royces and Yugos, now there's something in between," he says.
Some firms that in the past would have spent a lot of money creating their own data feed handlers and low-latency infrastructure or subscribing to consolidated feed services are turning toward hosted data consolidation and feeds to save money, says Mike Dunn, chief technology officer at Activ Financial, which has a hosted market data infrastructure that clients can access via TCP/IP connection. Yesterday, the company introduced an improved API that reduces latency between Activ's solution and third-party applications to less than 100 microseconds.
Market data vendors are launching new lines of business. This morning, Thomson Reuters announced an agreement with Smart Trade Technologies under which smartTrade liquidity management will be combined with Thomson Reuters Market Data System and Thomson Reuters' peer-to-peer messaging capabilities to provide clients with smart order routing and liquidity management. "We're seeing fragmented markets in every class, and clients are talking about the need to pull together the entire transaction workflow," Roche says. "We think this is a natural extension of our business; the most valuable data you can have is that you can execute on."
Also today, Bloomberg began offering a real-time market data feed from the Kazakhstan Stock Exchange through its Bloomberg Professional service. Bloomberg users can access real-time trading quotes, pricing and corporate information issued by the Kazakhstan exchange.
On Monday, Deutsche Boerse and Market News International rolled out an economic indicator news feed for algorithmic trading that provides employment, inflation, housing and other economic data from around the world, the organizations say, straight from the embargoed lock-up rooms from which such data originates.