This article provides an in-depth exploration of AWS Batch fair-share scheduling policies, explaining how parameters control resource allocation across different workloads.

• Fair-share scheduling meters equitable access to compute resources based on aggregate vCPU usage history
• Share decay time window (default 10 minutes) determines how long completed jobs influence resource allocation
• weightFactor parameter adjusts aggregate usage to give preferential treatment to specific shares
• computeReservation reserves capacity for inactive shares using formula: (computeReservation/100)^activeShareIds
• schedulingPriority parameter orders jobs within a single share, not across shares
• Fair-share queues smooth resource usage, enable dynamic allocation, and support mixed workload types
• Not suitable for jobs with special requirements like GPUs or large memory needs

Fair-share scheduling optimizes resource management in AWS Batch by balancing workload allocation over time while maintaining flexibility for priority jobs and new submissions.