Aria AI Operations Research Portfolio
Collection
Enterprise OR, optimization, and decomposition demos by Aria AI • 151 items • Updated
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
ClusterCast Scheduler Configuration
Configuration artifact for the ClusterCast GPU cluster workload optimizer.
Policies
| Policy | Objective |
|---|---|
max_gpu_utilization |
Pack GPUs densely via FFD, best-fit, backfilling |
min_wait_fairness |
Reduce queue time and tenant unfairness via DRF |
Algorithms
- First Fit Decreasing — Multi-dimensional bin packing heuristic
- Best Fit — Minimize node fragmentation
- Backfilling — Fill idle slots with preemptible jobs
- CP-SAT Gang — OR-Tools optional interval scheduling
- Dominant Resource Fairness — Multi-tenant fair share
- Min-Cost Flow — Greedy placement with transfer costs
- Rolling Horizon — Online re-scheduling windows
Constraints Modeled
Gang scheduling, GPU affinity, data locality, bandwidth, fragmentation, preemption cost, checkpoint overhead.
- Downloads last month
- 8
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support