Thomas Coleman and Shirish Chinchalkar
Department of Computer Science
Cornell University
USA
emails: {coleman, shirish}@tc.cornell.edu
Many of the important problems of risk management and financial engineering are computationally intensive. Computing good answers can take hours, sometimes days. Practitioners, under severe time pressure, sometimes make unwarranted assumptions or overly simplify models in order to compute an answer more quickly. Unfortunately, this approach can yield incorrect, and sometimes dangerous, computed answers.
The advent of (commodity) cluster computing, with point-and-click access from a desktop, offers convenient parallel computing power for the financial analyst or risk manager. We discuss some basic problem situations that arise in computational risk management such as portfolio management, pricing and hedging of large portfolios, Value-at-Risk, credit risk computations, and sensitivity analysis; we demonstrate the convenience and power of cluster approaches, especially when working at the portfolio level.