Use of distributed computing in derivative pricing

This paper compares two distributed computing environments when used to price financial contingent claims with Monte Carlo methods: a PC grid and a scientific computing Linux cluster. The paper also investigates the performances for different distributing strategies. On the basis of our experiments, a PC grid can be considered competitive with a scientific computing cluster. Both the cluster and the PC grid achieved nearly linear speed-up. We also find that it is optimal to set the number of jobs to twice the number of cores. Finally, we discuss the use of distributed computing in other fields of electronic finance.