SCIENCE

The Computational Finance Group of the Cornell Theory Center (CTC) has released a new online demonstration that illustrates an emerging approach to high-performance computing (HPC) and data management — XML-based Web services. Built on industry standards, Web services allow applications to interoperate across programming languages, platforms, and operating systems. Web services enable certain classes of HPC applications, specifically those that are very loosely-coupled, to distribute computation and data from a desktop or mobile device to remote servers or “workers.” Each “worker” solves only one part of the larger problem. Finance and life sciences are among the many fields with loosely-coupled applications that may be excellent candidates for Web services architectures. CTC chose an XML-based Web services architecture, Microsoft’s .NET Framework, to develop and test a solution to a high-performance computing problem in finance – the pricing of callable bond portfolios. These portfolios are particularly complex because they are “callable,” meaning the bonds may be redeemed by the issuer prior to the stated maturity date at the issuer’s discretion. Each individual bond is priced using a Monte Carlo simulation. This .NET Web-services approach to loosely-coupled HPC applications is demonstrated online by CTC at http://www.tc.cornell.edu/ctc-manhattan/callablebonddemo/. An Excel front-end launches the bond computations on a remote Windows cluster. Each bond is priced by one processor in the cluster, independently of all other bonds. As the bonds are priced, the results are returned to the desktop. “Up to 200 bonds can be priced in this demonstration with near-linear scalability. The important point, however, is that XML-based Web services architectures are extensible and reliable,” said Peter Mansfield, senior research associate for CTC. “Their extensibility means that adding servers to meet the performance requirement of the problem at hand is seamless.” “There are obvious advantages in ease of development and ease of deployment with Web-services technologies,” explained CTC Director Thomas Coleman. “For us, that means less coding and systems administration, and more time to focus on our core mission – enabling research.” While XML-based Web services makes sense for certain classes of loosely coupled HPC computations, Coleman believes that these technologies will further impact areas such as data intensive computing. “Moving data among geographically disperse, heterogeneous platforms requires industry-standard protocols such as XML and SOAP and secure interfaces,” Coleman emphasized. Cornell has demonstrated cross-platform data exchange using XML as part of its National Science Foundation Adaptive Software Project. CTC is a high-performance computing and interdisciplinary research center located on the Ithaca campus of Cornell University with additional offices in Manhattan. CTC currently operates an Intel/Windows cluster complex consisting of more than 1500 processors, in addition to Unisys ES7000 Servers. Scientific and engineering projects supported by CTC represent a vast variety of disciplines, including bioinformatics, behavioral and social sciences, business, computer science, engineering, finance, geosciences, mathematics, physical sciences. For CTC information, visit http://www.tc.cornell.edu. Information on computational finance is available at http://www.ctc-manhattan.com. Windows cluster information and training is available http://www.ctc-hpc.com.