Distributed Technology Techniques for Solving Dynamic Models

Solving large economic models requires large amounts of computational effort, as the complexity of these models increases the computational effort required in their solution increases dramatically. To examine the nature of these solutions researchers need to repeatedly solve models using different parameter sets, this compounds the need for computational effort. This paper examines the use of distributed computing as a way of providing large amounts of computational effort. It examines distributed computing projects such as “Seti@Home†which uses millions of computers supplied by volunteers to process recorded radio telescope data, the Distributed.NET project that deciphered the 64-bit RC5-64 cipher in 2002, the BOINC project that allows volunteers to specify the projects that their PC time can be used in. The paper proposes a technique that will use distributed computing to solve dynamic models. In their paper “Solving a saddlepath unstable model with complex eiganvalues†Herbert and Stemp point out that the solution of such models requires the use of a shooting algorithm and the choice of the correct solver, candidate solutions are chosen by a search algorithm and tested using the solver. A central server will use a database to log potential search areas and pass these on to the distributed computers who will then run algorithms to search for candidate solutions. Once complete the results of the search will be reported back to the server.