CastleNet's new Java-based global trading system is the first of its kind.

Although Sun Microsystems has long touted the benefits of Java programming and the ultra-thin client, much of these benefits have remained on the drawing board. Now, New York-based CastleNet, a fledgling software provider, is offering a Java-based trading analytic system capable of handling vast quantities of data and calculations in real time.

According to CastleNet founder Carl Carrie, the system grew out of a development initiative at New York-based brokerage Tullet and Tokyo. Currently, CastleNet is a privately held, independent company, although Tullet and Tokyo is its largest shareholder and, as of press time, its only client.

"The Beast” (a name coined by Carrie's son) can comfortably process 25,000 analytic transactions per second—that's 8 million such transactions per hour. Similarly, the system can support literally hundreds of open on-screen windows, each performing a different function.

So, how does the system achieve its impressive speed? According to Carrie, the system uses thin-client architecture to optimize system resources. At Tullet, software and data reside on nine servers in New York, as well as several additional servers in London and Toronto. Load-balancing middleware allows tasks to be efficiently distributed among these servers on a global basis. Because most off-the-shelf middleware packages were too slow, CastleNet designed its own proprietary middleware package that manages computing resources as well as the traffic of messages throughout the system.

Putting all transactions on the server level reduces the burdens on both the users' client machine and the company network. The original Tullet system was written in Visual C++, and the latest version is 100 percent Java. Both systems, however, allow users equipped with a web browser to access the system—and retrieve information and graphics in real time. "This is not just a thin client, it's an anorexic client,” says Carrie. "The system can run on Java stations, Network computers, [even] pared-down PCs.”

This configuration has a number of major advantages. First, if one of the servers goes down, the middleware will simply reallocate resources among the remaining servers. The event would thus go unnoticed by traders and allow them to continue business activities without disruption. Second, because all critical software resides on the servers, it is easy to update functionality without having to manage upgrades across what could be hundreds of client machines. Third, because client requirements are so minimal, the system does not require an enormous investment in new hardware.

Carrie stresses, however, that the Beast is not an "Internet” system. Tullet does not access the system through public Internet lines, nor is the system intended for that sort of access. Instead, he says, Tullet allows remote access to the system through dedicated phone lines. "This system is meant to be run on secure intranets or extranets, not through the public domain,” he emphasizes.

But what does it actually do? Basically, there are three components to the Beast's functionality, the most important being to provide sophisticated deal analysis. So far, Tullet uses the system for pricing interest rate transactions, ranging from plain vanilla to extremely complex. The system can accommodate multifactor models, such as those from Hull and White, and also includes scenario analysis. All these analytics are performed in real time. Currently, however, the Beast is built for analysis only; it is not a deal-capture system, and it does not include any sort of portfolio analytics. "But,” says Carrie, "we could modify the system to store and handle actual transactions without a great deal of effort.”

It also provides market data and news reports from a wide variety of sources, including the Internet. CastleNet can provide some data directly, as well as real-time links to Reuters, Telerate and other popular sources. Users can also custom-tailor their news reports, so they get consolidated reports, constantly updated, from multiple sources, displayed in a single location.

The final piece is a corporate communications package, providing a gateway for e-mail and, more critically, allowing users to share information in real time, including documents, analyses and so on. At Tullet, for example, brokers input current bids and offers, which then become instantly available to other brokers worldwide. Likewise, multiple users can review and comment on the same reports.

Summit on NT Some say that the graffiti is already on the wall, that it is only a matter of time before Windows NT becomes the platform of choice, even within the UNIX stronghold that is Wall Street and even for heavy-duty transaction processing applications. In anticipation of this brave new world, Summit Systems has released its product on the Windows NT platform, and it is now compatible with Microsoft's SQL Server database. The current Summit production release (Version 2.5) is fully supported on Windows NT (3.51 and 4.0). All existing UNIX features and functionalities have been implemented on Windows NT, including Summit's API and extendibility tool kits. Many experts believe that the movement of a major, "industrial strength” trading and risk management system such as Summit to the NT platform may signal an industry trend.

Carrie explains that the user interface is "ergonomically designed,” and users can easily customize their screens. He says, "We hired a designer who had worked on Jurassic Park to help design our screens.” Indeed, it does seem fitting that the Beast received help from one of the minds behind creating images of giant dinosaurs. Certainly, as far as enterprise-wide, Java-based financial applications go, the Beast is, at least for now, the 800-pound gorilla in the field.

For more information see www.thebeast.com.

Adds to FEA's Add-in

Most software is a work in progress and, even if a product has been on the market for a long time, there may still be some new tricks to teach these old, faithful dogs. The upgrade version of Berkeley, Calif.-based Financial Engineering Associates' new release of @INTEREST, which is part of FEA's popular collection of low-cost, high-functionality spreadsheet add-ins, includes some important new yield curve features.

"The customer analysis process usually begins with the extraction of a zero yield curve, and so we introduced our new ZCURVE functionality to provide a customized best-fit to traded market instruments,” says FEA president Mark Garman. "We also introduced fast, new calibration functions that use the volatility information in traded option prices to analyze, in turn, more complex derivative instruments.” The next step was to provide clients with a wider range of derivatives analytics, such as American swap options, and Bermuda swap and bond options with irregular strike prices and "captions” and "floortions.”

Comply, Now! Los Angeles-based Capital Management Sciences has released an enhanced compliance tool. It's BondEdge product, a high-level, fixed-income analytical software package aimed at investment managers, now includes a new feature designed to keep unruly traders in line. Released at the close of June, this compliance features gives users a reliable way to determine whether or not any proposed transaction is in compliance with investment policy guidelines. Portfolio managers can use the Compliance System to define dynamic investment guidelines for all sorts of fixed-income products, down to the issuer level. Any transaction that would take the portfolio out of compliance would be flagged in "pre-trade” stage, before any sort of deal has been executed. According to Teri Geske, director of product development for CMS, "Compliance with guidelines and regulations is becoming increasingly important to portfolio managers, especially as portfolios grow more complex.” With the new CMS product, portfolio managers will be able to keep complexity under tight control.

Although the FEA software has always supported a wide range of models, it has been difficult for users to put the most modern models to work because their volatility parameters were not quoted in comparable terms. The latest version of @INTEREST, which Garman calls "a mini-breakthrough,” places all of these models on a comparable term-volatility-quotation basis. Now all @INTEREST volatilities conform to historical market practice—that is, the older Black model—even when the later, more sophisticated models are used. "This step seems small,” says Garman, "But it may actually be quite important in stripping away barriers to using more advanced models within the interest rate derivatives markets.”