Priority based queue

To achieve the optimal usage of the cluster, we will operate under a priority model by which compute resources of owners not in use can be scheduled and used by other investigators. Owners will have highest priority for the number of nodes they have purchased in the job queue. With the priority model, the central management team will schedule these resources for optimal overall use, in addition to maintaining the hardware. This will result in a lower cost of operation for owners.

The Priority Service Model can be made further granular by adding more checks and balances; for example, to ensure owners are not always using these resources for lower costs, we can associate a fixed number of compute cycles per year to an owner. As long as the owner works at less than that threshold, their tasks are automatically top priority in the queue. If that limit is crossed, priority will be reduced and the owners’ tasks will take turns with all other tasks with the same priority level. With correct scheduling mechanisms and capable infrastructure, the priority model can be great success.