Parallel computing using idle workstations
Jiubin Ju, Gaochao Xu, Jie Tao · ACM SIGOPS Operating Systems Review · 1993
Enormous aggregate computing capacity in distributed computing systems composed of workstations connected by high speed local area networks may exceed that supercomputers often have. The Sidle (for Sharing Idle Workstatinos), a system for scheduling idle workstations and remote executing which is currently running on 19 SUN4 workstations connected by an Ethernet is presented in this paper. Sidle enables user's many programs or sections of a program to be executed parallelly on many idle workstations. Sidle has many features compared with existing systems: It allows more than one job to be run on an idle workstation and nested remote executing, makes use of CPU utilization much better, with more higher transparency to the users and better performance.