Extending the computational bandwidth of engineering workstations

The authors describe their experience so far with the development of a distributed facility called REM (remote execution manager), which supports load sharing and remote execution in a local network of workstations. REM is implemented at the application layer and runs on workstations running Berkeley Unix. The authors describe the motivation and justification behind the development of REM and given an overview of REM as it exists now. The load sharing is described, and the interprocessor communication overheads are compared with similar systems. The methods used to achieve fault tolerance in case a workstation crashes or is withdrawn from the pool of servers are outlined. Some of the preliminary results and plans for future work are described