Job Scheduling For The Bluegene/L System

Cellular architectures with a toroidal interconnect are effective at producing highly scalable computing systems, but typically require job partitions to be both rectangular and contiguous. These restrictions introduce fragmentation issues which reduce system utilization while increasing job wait time and slowdown. We propose to solve these problems for the BlueGene/L system through scheduling algorithms that augment a baseline first come first serve (FCFS) scheduler. Our analysis of simulation results shows that migration and backfilling techniques lead to better system performance.