SIFMA Conference puts spotlight on Wall Street’s need for speed.
Securities trading in the financial-services industry is increasingly becoming a race to be the fastest to electronically complete a deal between buyer and seller, driving the adoption of automated trading and a demand for ever-higher network speeds.
With that in mind, a group of high-tech vendors at the Securities Industry and Financial Markets Association’s (SIFMA) 2007 Technology Management Conference and Exhibit on Tuesday plans to unveil their latest products and services designed to address Wall Street’s ravenous need for speed in an era when a few milliseconds can mean winning or losing millions.
In their own bid for victory, Voltaire, Intel and HP have joined forces with Reuters, which provides the financial-services industry with its Reuters Market Data System, to come up with a high-speed, low-latency network architecture they claim traders should adopt. RMDS is installed at about 2,500 financial-services firms.
“Speed is the key to competitive advantage, whether it’s from the trading desk or automated,” said Anne Ambrose, HP’s director of brokerage trading. “Otherwise, you won’t be able to participate in the trade.”
Reuters business manager Nicolas Pintart said his company worked to ensure the RMDS middleware, which analyzes market data as it’s flowing from various financial and news sources, has been optimized to work in combination with Voltaire’s 10-20Gpbs Grid Director InfiniBand-based switches and Remote Direct Access Memory technology, HP ProLiant BladeSystem c-Class servers, Intel’s Quad Core Xeon 3500 processors.
Patrick Guay, senior vice president of marketing at Voltaire, said the collaborators started a year ago to engineer an architecture for the financial-services industry.
Nigel Woodward, director of financial-services solutions marketing at Intel, added the technology model that HP, Intel, Reuters and Voltaire devised is presented as an approach that financial services can adopt without having to go through extensive engineering analysis on their own.
Tom Price, senior research analyst at TowerGroup’s Securities and Investment Practice based in Needham, Mass., notes that Bloomberg, Reuters and Thomson Financial (which is in the process of acquiring Reuters) compete to take feeds from stock-exchange and news sources and sell that data to brokers and dealers.
He pointed out that the financial-services industry is becoming increasingly dependent on automated trading — known in the industry as algorithmic trading or simply “algo trading” — to complete trades. In fact, machine-to-machine algo trading is generally assumed to account for about 60% of electronic trading.
“Basically, you have parallel infrastructures today,” said Price. “One is humans trading, the other is machines trading.”
Price said that algo trading relies on computer applications to make buying decisions based on complex programs built around algorithmic logic, and it was pioneered by firms such as Goldman Sachs, Merrill Lynch and Morgan Stanley as far back as a decade ago. Today it’s become the industry’s driver, with hedge-fund activity believed to rely primarily on machine-based algo trading.
“That machine is sending out orders and canceling out orders, and it’s happening very fast,” Price says. “A millisecond matters because it’s about the need for speed, and whoever gets there first wins.”
Intel competitor AMD has partnered with Cisco and its InfiniBand-based switches to introduce systems for the financial industry, and AMD’s senior manager Richard Reichgut agrees algo trading is driving speed requirements down to the millisecond, especially as there is an “the exploding volume of data right now.” He said the concept is to provide data from the network directly into memory of a system without hitting the CPU.
Varghese Thomas, vice president and general manager of financial markets at SAVVIS, which hosts five major exchanges at its New Jersey-based data center, said co-locating the Reuters, Bloomberg and Thomson Financial servers as close to possible to stock-exchange electronic communication network servers can help reduce latency and gain a few milliseconds of advantage.
But he noted that some financial-services clients would like to bypass the market-data aggregators, such as Reuters and Bloomberg, and go directly to an exchange on their own circuits, if only to save time.
“Everyone is working on high-performance computing, and the application development folks say the network has to be better, and the network folks say the applications have to be better,” Thomas noted. One thing is growing more obvious, he adds: Automated trading is taking over because “humans can’t keep up with it.”
Thomson Financial earlier this year teamed with BT Radianz to provide what it calls a low-latency Thomson Data Feed, and Matthew Burkeley, Thomson’s head of strategy, agrees that now “everyone is obsessed with speed.”
Today, the financial-news reporter is sometimes regarded as a bottleneck slowing critical data that could affect the market from reaching traders that need to know it, Burkeley points out.
Thomson Financial generates almost 2,000 stories a day completely through computer generation that doesn’t require a human writer, and now both Reuters and Thomson are individually coming up with ways that machines can automatically read and rate the criticality of news delivered electronically and integrate that information directly into algo trading without a human reader having to be involved.
Naturally, vendors have different ideas about the best way to optimize speed and lower latency in trading. One vendor exhibiting at SIFMA, GigaSpaces, will be talking up its software-based “Extreme Transaction Processing and Real-Time Analytics” for the automated trading environment.
According to Geva Perry, chief marketing office, the GigaSpaces eXtreme Application Platform Version 6.0 is middleware that functions as an “in-memory data grid” that replaces the traditional role of databases by combining and processing trading data on the fly without having to rely on a back-end database, Perry says.
“The database is no longer the repository for the real-time truth, it’s just an archive,” he says.
The latest version of the GigaSpaces eXtreme Application Platform, expected out in September, uses blade processing and adds support for the programming model called “Plain Old Java Objects” and the open source Spring Framework.
GigaSpaces believes provides more processing efficiency over the older Java Beans technology. eXtreme also supports plain .Net and plain C++ objects. In addition, GigaSpaces is also announcing a partnership with Microsoft to make it easy to use the Microsoft Excel database as the front-end user interface for the GigaSpaces in –memory data grid. The GigaSpaces software is used today by the Chicago Mercantile Exchange, Dow Jones and Societe Generale, among others.
While trying to wring out greater speed in trading is important, the financial-services industry is pondering other technology questions, such as whether to adopt VoIP.
Siemens today announced the HiPath Trading System 4.0 for voice and data communications, introducing a VoIP-based hardware-based “turret” for traders’ voice communications called the High End Communications Terminal Open Range (HECTOR).
Expected to under $10,000 per seat, HECTOR has a touch screen and Windows XP-based clients to combine voice, e-mail, videoconferencing and instant messaging. Don Lisco, Siemens’ director of sales for the financial sector, said HECTOR is essentially a Web appliance with VoIP communications based on the Session Initiation Protocol.
The HiPath Trading System 4.0, expected to ship in November, will also support legacy Time-Division Multiplexing terminals
On the trading turret front, BT today announced the Virtual Integrated Trading System, which works like the traditional turret at a dealer’s desk used to support voice and data communications but will be supported as a managed service in the New York City area.
Phil Mottram, BT’s managing director of trading systems, says the switching for Virtual ITS will be supported through switches directly in BT’s New York City-area data center with connection to the BT Radianz network for the financial-services industry.
“With Virtual ITS, we’re targeting the hedge funds in the New York tri-state area, particularly the smaller ones,” Mottram said.
