Using Java, Java-enabled Web and object-oriented programming technologies, a framework is designed to organize multicomputer system on Intranet quickly to complete Monte Carlo simulation parallelizing. The high-performance computing environment is embedded in Web server so it can be accessed more easily. Adaptive parallelism and eager scheduling algorithm are used to realize load balancing, parallel processing and system fault-tolerance. Independent sequence pseudo-random number generator schemes to keep the parallel simulation availability. Three kinds of stock option pricing models as instances, ideal speedup and pricing results obtained on test bed. Now, as a Web service, a high-performance financial derivative security-pricing platform is set up for training and studying. The framework can also be used to develop other SPMD (single procedure multiple data) application. Robustness is still a major problem for further research.
展开▼
