Distributed computing with personal computers

✍ Scribed by Christophe A. Poulain; Bruce A. Finlayson

• Publisher: American Institute of Chemical Engineers
• Year: 1996
• Tongue: English
• Weight: 484 KB
• Volume: 42
• Category: Article
• ISSN: 0001-1541
• DOI: 10.1002/aic.690420128

✦ Synopsis

Computer programs for blood-tissue exchange (BTEX) modeling allow one to fit experimental data, find parameter estimates, and perform a sensitivity analysis. However, such extensive computations must be rapidly carried out to be practical.

To date, we have focused on the numerical technique used to solve the BTEX modcl equations since the most efficient algorithm is essential to carry out the calculation quickly . Nevertheless, hardware considerations are important. While supercomputers and mainframes are the fastest, they are very expensive; in recent years attention has focused on harnessing the computing power of idle workstations or personal computers distributed on networks Sharma et al., 1991;. This approach is attractive, because it can provide large computational resources at a minimal cost. The Dept. of Chemical Engineering at the University of Washington has a cluster of networked Power PC Macintosh computers. This resource prompted us to implement a distributed version of a multipathway blood-tissue exchange model. In doing so, our overall objective is to demonstrate that parallel computation is an efficient approach to decrease computational times in blood-tissue exchange modeling, and that it can be done by scicntists using standard tools. This approach is also applicable to other problems such as: stochastic (Laso, 1994) and Monte-Carlo simulations in which a model equation containing a random variable is solved repeatedly; the calculation of sensitivity coefficients of a model ; and parameter estimation.