Reproduce “Solver-Free and Cost-Aware Online Scheduling of AI Data Center Tasks” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Xin Qin, Dawei Qiu, Fei Teng

The authors propose a per-arrival rule that selects the cheapest task start fitting GPU and CPU capacity while reserving room for high-priority tasks. In a 26-hour simulation, flexible-task cost was $340.74, versus $341.27 for a five-minute rolling MILP; the rule reported no late high-priority tasks. This result depends on price forecasts and sufficient capacity for running and high-priority tasks.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.