# Allen Institute rolls out budget‑driven GPU scheduler

> The AI Infrastructure team replaced its priority system with a budget‑based fair‑share scheduler, letting researchers claim a share of GPU time and cutting on‑call repairs by 74%.

Oossa · 2026-10-09 · https://oossa.com/en/allen-institute-rolls-out-budget-driven-gpu-scheduler

Allen Institute for AI (Ai2) introduced a new GPU cluster scheduler that uses budget allocations and hierarchical fair‑share to decide which research jobs run. The system gives each project a fixed share of GPU hours – for example, Project A1 gets 35% of total capacity – and only jobs backed by a budget are protected from preemption. The scheduler also requires a minimum runtime contract, which lets it reclaim resources for other jobs after that time. Engineers say the change feels like a 30% boost in usable compute and reduced manual repair work by 74%.

## The facts

- The blog post was published on 2026-10-09.
- Ai2 manages clusters of 88‑1024 GPUs serving about 150 internal researchers.

## Why it matters

Researchers now get a predictable share of GPU time and on‑call staff spend far less time manually shutting down long‑running jobs.

## Sources & references

1. [Impactful scheduling for GPU clusters](https://huggingface.co/blog/allenai/impactful-scheduling) – Hugging Face, 2026-10-09
2. [Impactful scheduling for GPU clusters](https://allenai.org/blog/impactful-scheduling) – Ai2

Last updated: 2026-10-09