Compute and entitlements
Aidin runs your workloads on shared platform compute. This page describes the allocation model at the customer level: what you get automatically, what needs an entitlement, and how contention is handled.
The allocation model
Section titled “The allocation model”| Resource | How you get it |
|---|---|
| CPU and memory | Included with your workspace, bounded by your project quota |
| Workspace storage | Included with your workspace, bounded by a storage budget |
| GPU | Requires a GPU entitlement on your account — not granted automatically |
| Reserved / dedicated GPU capacity | Enterprise agreements only (limited access) |
Entitlements
Section titled “Entitlements”An entitlement is an account-level grant (from your plan, quota allocation, or enterprise agreement) that unlocks a resource such as GPU. Entitlements are managed by Aidin, are visible on your account, and cannot be self-granted by editing workspace settings.
During the public launch, GPU entitlements are not offered self-service; see the launch limitation note on Compute profiles.
Fair-share scheduling
Section titled “Fair-share scheduling”When demand exceeds capacity, workloads queue and are admitted fairly (queue-aware, priority-aware scheduling) rather than failing. Practical consequences:
- A workload may start shortly after you launch it if capacity is free, or wait its turn during peaks.
- Stopped notebooks immediately release their CPU/GPU reservation; nothing is reserved for stopped workloads.
- Your quota is yours: one project cannot consume another project’s quota.
See Admission and lifecycle for the scheduling behavior in detail.
Per-plan numbers
Section titled “Per-plan numbers”Exact CPU/RAM/GPU quota numbers per plan are being finalized with the public capacity and billing model and will be published on Storage and quotas together with Billing. No numeric quota is documented here before it is approved.