“It’s an arms race.”

That’s how Dr. James Holt describes the effort that Chicago-based Townsend Analytics, Ltd. confronts in keeping ahead of the fast-growing volumes of quotes and messages coming from exchanges, ECNs, and other liquidity providers. Customers of the firm’s RealTick® market data, analysis, and direct access trading platform include retail investors, day traders, hedge funds, money managers, mutual funds and sell-side broker dealers. It is a client base that has little tolerance for slow performance. Many customers, who are across the United States and in 93 other countries and territories, notice immediately if the company provides less than sub-second response time. The low end of recommended connectivity is a DSL or cable modem via the Internet, and on the high end, clients are running direct lines, or accessing RealTick through third-party networks like Radianz. The clients typically are running fairly powerful PCs – they need at least a one gigahertz processor to keep up with the price feeds on fast-moving stocks since some stocks generate more than 500 price quotes per second. Multiply that by a large watch list of symbols, and that quickly adds up. Holt doesn’t expect market volumes or speeds to slow down anytime soon.

Dr. James Holt finds financial markets technology challenges similar to the particle physics work he did at the Fermi lab.

“If you offer more capacity, people want it.” In the future, he thinks the industry will have to find better ways to represent what is happening in the markets. That might be a business intelligence box that can assimilate the data and present it in a coherent fashion so a trader can make intelligent decisions, or perhaps some new form of visualization.

Founded in 1985, Townsend Analytics provides tools and support services for both institutional and retail brokerage firms for multi-route trading and display of real-time quotes, analytics, risk analysis, and other applications including televised displays of streaming data.

Because it offers a comprehensive range of products, including equities, fixed income, foreign exchange, futures and options, Townsend Analytics provides a single platform that traders can use for complex trades across multiple asset classes. Traders can also consolidate all their various accounts (by asset class, currency, broker) by creating virtual accounts. Today, when many of the direct access firms and trading software companies have been acquired by banks or brokerages, Townsend Analytics remains one of the last true independent and broker-neutral technology vendors.

Unlike some market observers who think computers will take people out of trading, Holt expects humans will remain important players.

“There will always be someone who can look at a trend or a subset of data and make money from it, because the market is never 100 percent efficient,” he said.

That’s not for lack of speed and volumes of data, which have both been increasing relentlessly, forcing companies like Townsend Analytics to run in order to keep up.

“We are constantly improving our software to optimize speed and bandwidth,” said Holt. “While RealTick has periodic product releases, on the server side the improvements are continuous. Close to a third of the employees are developers who are adding features and looking for ways to speed up delivery.” Unlike many companies, at Townsend Analytics, the development and operations staff work together closely.

“All of management is technically sophisticated,” added Holt. Some of founder Stuart Townsend’s original code is still running in the analytical engine, and Stuart is widely acknowledged as both an expert in the financial services industry and as a technical architect.

“Senior management here understands both the business and the technology. They don’t come into meetings with ideas they have picked up from reading airline magazines. Bad ideas get shot down fast,” Holt said.

To stay ahead, Townsend has replaced its c-tree database with SQL Server 2005 to process over 90 million equity quotes per day, store prices in the database and generate charts and analytics from that information. The company is using a beta, or pre-release, version of the new database in full production in an application that’s about as mission critical as it gets in trading analytics.

“SQL is about the most stable thing around here,” said Holt. Townsend Analytics has been testing the new release for more than a year and has been in production with it for several months.

Townsend Analytics has partnered with Microsoft under its Premier Partner Services program. In return for putting the database through demanding trials under the supervision of an expert staff of technologists – Holt came to Townsend from the Fermi Labs – the trading technology firm has access to the top technologists at Microsoft.

“It’s a wonderful arrangement, because it gives you an advocate at Microsoft,” Holt explained. “Often we are exploring new technology and need new information. They will arrange calls to experts within Microsoft for us. I like working with Microsoft because we are both pushing the edge on software.”

Moving from Fermi’s particle accelerator to working in financial markets didn’t faze Holt.

“The data flow of a particle physics experiment is very similar to the data flow in our industry,” he explained. “You have a high event rate, and then you have triggers off that high event rate. Both require real-time analysis of the events, and then you have to both store and route the results. It requires real-time monitoring to make sure the operations are going correctly.”

Holt finds that the SQL Server 2005 database offers significant improvements over the previous version.

The tests his group has run demonstrate that it is fully 20 percent faster. Integration services have been improved and users can link the database more easily to applications written in .NET.

“It is much easier to use .NET with SQL Server 2005,” said Holt.

Townsend Analytics works closely with Microsoft developers to provide a highly demanding test site for new applications.

“We spend considerable efforts testing code at high speeds,” said Holt. Townsend Analytics developers have run tests where Microsoft’s C# programming language has outperformed C++. “And C# is easier to program,” he added.

Because Microsoft’s development tools are both powerful and easy to use, developers can test new concepts, such as ideas for using data mining with RealTick data, quickly. “If you can dream it up, you can incorporate it into the product,” explained Holt. One new feature is Slicer, which is a native algorithmic trading feature to slice up large orders into smaller increments quickly that can be fed into the market at intervals that are calibrated to avoid large price movements and help traders achieve a better VWAP on their trades (volume-weighted average price). Also new in 8.0, users can create their own customized technical studies for trade opportunity analysis or for triggering automated conditional orders, based on their custom criteria. Charts also now include visual representations of trades and orders, along with color rules and other indicators showing when various conditions have been met on a historical basis.

Townsend Analytics runs its analytics on a scale-out model – lots of high-end commodity servers rather than a few very big machines. While a larger group of servers can present management problems to teams that aren’t familiar with them, the scale-out model offers some significant advantages. A firm can add relatively cheap servers one at a time, instead of expanding capacity in $100,000 or $500,000 increments.

“These are not your grandmother’s PCs,” added Holt. The company runs two- and four-processor machines from the popular leading hardware providers including AMD Opterons at the high end. SQL Server does an excellent job at operating across a large number of servers, he added.

To keep ahead of rising data demands, the Townsend team uses every trick in the book on a multi-threaded, multi-server environment to extract top performance from its servers, he said.

Trend Watchers

See the news before it happens - Hottrend uses sophisticated monitoring to detect subtle changes in trading that can presage price shifts.

Michael Felix, director of marketing at Townsend Analytics, Ltd., joined when the company acquired Hottrend, a firm he had co-founded along with Jeff Brown and Luc Delfau to spot abnormal activity in the stock market.

Hottrend monitors all equity securities in real-time and detects changing trading behavior patterns through a sophisticated multivariate model. The system compares each security's trading behavior to its historical expectations and alerts you when “mindshare” around a security is changing as the result of new information being disseminated throughout the market.

By looking at the number of trades, volume, rate or price change, and other indicators, it points users to liquidity events as they begin to occur, said Felix, who describes it as a tool for both trade opportunity discovery and for risk management. By spotting changes in a complex array of data around each stock, it can provide an early indicator that something is happening, and alert traders who can then check news and other sources for more information in order to make a trading decision. With over 10,000 financial news stories moving across the wires each day, no trader can possibly keep track of all the news on the wires, so they need to use news in conjunction with a service such as Hottrend.

“Because the risk management implications for Hottrend are huge, we needed a way to provide a broader set of customers with an alerting mechanism, wherever they happen to be,” explained Stuart Townsend, president of Townsend Analytics, Ltd. “A fund manager holding securities might not have enough time to track all the tick-by-tick quotes or news stories that relate to them. But, if something abnormal is occurring with a security, no matter what the driving force – news, rumor, leaks, or other activity – the fund manager must be decision-enabled as soon as possible in order to take decisive and safe action. Hottrend Notification Services acts as an information accelerator for the fund manager.”

Townsend Analytics used Microsoft SQL Server Notification Services to build the securities alert program. The simple programming framework and scalable server solution based on Microsoft SQL Server 2000 allowed the company to quickly develop an innovative notification application efficiently.

Traders and investors access Hottrend Notification Services through RealTick. First, they must identify all the securities they want Hottrend to watch. They can easily populate symbols in the Hottrend Notification Services integrated Web browser interface in RealTick using a drag-and-drop from their existing watch lists and portfolios, or they can enter the symbols directly. Then, for each security, they customize how and when they want to receive alerts.

“When traders are sitting in front of the computer, there's a great deal of information at their disposal – data that is both fundamental and technical,” said Felix. “We've tried to consolidate by feeding information from different sources into the Hottrend engine and providing our customers with an alert that says, ‘this is the security you should be looking at now.’ Ultimately this consolidation of relevant information frees up traders' time and helps them to be more efficient.”

Free online classes in how to use RealTick and Hottrend occur on a regular basis. See the company’s Web site for more information – www.realtick.com.