Moving beyond the hype, Roger Lang, director, marketing and development at NYC’s Cornell Theory Center, shows how Web services will underpin explosive growth of high-performance computing in financial applications.

An increasingly complex and competitive global trading environment, shortened product development cycles, enormous pressure on costs – these are some of the conditions defining what we may as well call reality financial services.

With regulators, clients and investors clamoring for increased amounts of information, business and IT managers are forced to achieve the highest possible levels of operational efficiency while optimizing available infrastructure.

Help is on the way.  Once-upon-a-time Buck Rogers-type developments in the lab are now ready for enterprise deployment. Early adoption may prove to be a critical step forward for firms large and small, buy-side and sell-side, eager to maintain competitiveness, particularly in the areas of trading and risk, as well as highly quantitative data analysis.

Many of these developments come from what is now referred to as high-performance computing (HPC).  For decades, the approach to HPC was the same. Specialized equipment and software requiring advanced parallel software programming expertise were a constant, inhibiting the technology’s move from scientists’ hands to the commercial arena.  

That’s changing with staggering results.  Since the late 1990s, the Cornell Theory Center (CTC), a global leader in high-performance computing, has demonstrated the value of making the transition from UNIX-based proprietary architectures to industry-standard HPC and the power of .NET for customers in finance, bioinformatics and other industries. CTC High-Performance Solutions, which provides complete, enterprise-level custom solutions for Windows-based HPC and .NET applications, is backed by an alliance of industry leaders, including Microsoft and Intel and the CTC, with support from Unisys. CTC-Manhattan, a satellite facility located at 55 Broad Street in Manhattan, is a Windows HPC showcase with a special emphasis on computational finance and .NET solutions.

As Tom Groenfeldt, editor of Windows in Financial Services wrote in 2003, “The CTC has proven that clusters of Windows servers and PCs can not only reduce TCO and provide cost savings to the IT department, but more importantly empower all levels of the organization with cluster-ready applications that scale on demand.”

For example, a high-performance cluster computer using Windows Server 2003 and messaging software, prices a portfolio of complex derivatives by breaking up the calculations and distributing them to multiple remote processors, with the answer delivered to the desktop in minutes, not the eight hours of overnight processing it once took.

Today, pilots running on Excel are using Web services to broadcast trading analytics to an array of servers and getting answers in just seconds.  

HPC’s new roadmap using industry standard solutions from Microsoft and Intel has begun transforming the science of parallel high-performance computing into an everyday business application, transparent to the user. Linked to computational steering to harness the power of databases and drive data elements out for analysis, these everyday tools are transforming technology once restricted to the lab into a driving force in the marketplace.  

The CTC’s computational finance group led by Professor Thomas Coleman has designed practical solutions for complex mathematical and computational problems for the financial industry.  With experts in interest rate modeling, bond pricing, derivatives pricing, Monte Carlo simulation, intensive computing and software design and implementation, his team offers expertise in large-scale portfolio optimization, dynamic hedging, pricing exotics, volatility modeling and the valuation of sophisticated insurance products.

Structured bonds

A large Asian bank wanted to improve methods to price complex structured bonds with embedded options that allow issuers to call them in before maturity. CTC’s engineers priced portfolios of these bonds by implementing a technique called Least Squares Monte Carlo (LSM). Future interest rate paths are simulated. In many cases it’s optimal to call in structured bonds before maturity.

Convertible bonds

To accurately measure and manage the risks of holding convertible bonds, FSIs must consider various potential scenarios. Because of the variables involved, these scenarios generate an extremely large data set.  Generally, calculating the data isn’t feasible on one workstation. It is best handled with parallel computing and Web services. To solve this risk management problem efficiently, CTC simultaneously runs multiple copies of software developed by ITO33, a French-based company, to price convertibles under exceptionally large sets of different user-defined scenarios. The Web services solution is run on a Windows cluster accessed and managed by Microsoft’s .NET platform. Figure 1 shows the relationship between the desktop computer and the HPC clusters.
 
Risk Management

The risk management solution in Figure 2 includes an easy-to-use graphical interface that allows firms to examine the sensitivity of future convertible bond prices to changes in the variables that impact price. With a few mouse-clicks, different types of graphs can be built to study these sensitivities.  In addition, worst case scenarios are identified automatically and highlighted for management review and discussion.
 
Defaultable bonds

Although attractive to bond portfolio managers because of their yields, corporate bonds have their own inherent risks – yields are higher because corporate bonds are risky and their values are impacted by market risks and the possibility that the issuer will default. CTC’s solution to pricing portfolios of risky bonds and calculating their VAR and conditional VAR uses an Excel interface on the desktop. It sets interest rate model parameters; sets the default parameters for each of the bond-issuing corporations; builds up to 10 portfolios of interest; and selects the desired number of simulations of future interest rate paths.

The Excel interface accesses a cluster running Windows .NET server, gathers the results and displays them for review.  The interface accesses a cluster running Windows .NET server, gathers and displays results. Figure 3 shows program control buttons, performance and efficiency monitoring, selection of portfolios included in the computations and numerical and graphical results of portfolio computations.
 
Live HPC Cluster Demonstration

If you would like to use one of the CTC’s clusters from your desktop, go to www.ctc-manhattan.com, click on the Live Demonstration link and follow the instructions. Editor’s Note: Excel 2003 Professional Edition is needed.

The new paradigm: HPC meets .NET and Web services  

All of these examples point to a new HPC roadmap that is emerging based on Web services.  The CTC is delivering a wide range of finance applications that utilize .NET and Web services to access remote clusters, applications that often utilize Excel, serving as the front end.  

HPC has moved beyond the classroom and the lab. Now it’s ready to enter trading rooms around the world.   

That’s no hype … that’s simply reality financial services.

For more information about Cornell Theory Center, visit www.ctc-hpc.com.