Scheduling jobs with different CPU and memory requirements often leads to resources being stranded or underutilized across at least one dimension e.g., CPU or memory. This leads to sub-optimal utilization of data center or cloud infrastructure and wastage of valuable resources. Solutions to address the resource stranding problem that involve pooling resources across different machines are complex and impose extra hardware requirements. This disclosure describes a shape matching optimization technique that is a flexible software-based solution to preferentially allocate jobs to machines by matching the shape of the job to the available shape of the machine. The shape matching scheduler improves resource utilization.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.