Distributed Computing with RESTful Web Services

High performance applications such as gene sequence alignment searching, data mining, graphics rendering and Monte Carlo simulations in finance and science are putting ever-increasing load on computing infrastructure. Traditionally, such pleasingly parallel workloads are parallelized using standard batch scheduling or, on a larger scale, web-service-based grid middleware, which is traditionally based on SOAP web services. It is argued that Restful web services provide a natural framework for such asynchronous distributed computing and allow an architecture more suited for on-demand computing. Concretely, the services of the Grid Factory distributed computing system are described. These services are implemented as Apache modules, leveraging Apache APIs for access control and database access. Moreover, the standard Apache modules mod_dav and mod_ssl are used for file serving, authentication and and encryption. The characteristics of the system are described and the performance of this direct use of Apache as hosting platform is evaluated. The system has been subjected to initial testing, demonstrating the feasibility of the approach.