| With the advent of the Internet, networks of workstations can be used to form a metacomputer for running coarse grain parallel applications. The advantages of metacomputers over specialized massively parallel computers are wide availability and cost-effectiveness. Long-running computation on metacomputers, however, requires both fault tolerance and the effective utilization of idle workstations.; This dissertation reviews existing approaches to create a metacomputer, and then presents the design and implementation of a new model called Anaconda based on the Java language and using a virtual shared memory to distribute work among idle workstations. The main features of this new model are adaptive parallelism and fault tolerance . With adaptive parallelism, programs execute on a pool of workstations that dynamically grows and shrinks in response to both the availability of idle machines and the availability of parallelism within the program. The model provides transparent fault tolerance. Experimental results are also presented. |
