COMPASS-ABS: Reducing Fragmentation in Shared GPU Clusters for Deep Learning Training Workloads

arXiv:2609.18519 · cs.DC, cs.LG · Submitted 2026-09-16 · Read on arXiv

cs.DC, cs.LG

Submitted: 2026-09-16

Updated: 2026-09-16

Comments: 22 pages, 7 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Trace — 34× in this paper · explained in 2 episodes
• Anchor-Based Space — 8× in this paper
• Scheduler-Induced Fragmentation — 8× in this paper
• Alignment — 6× in this paper · explained in 1 episode
• Robustness — 6× in this paper · explained in 5 episodes
• Under the Workload Composition Condition — 4× in this paper
• Theoretical Guarantees — 2× in this paper · explained in 2 episodes
• Transformer — 2× in this paper · explained in 2 episodes
• Abstraction — 1× in this paper · explained in 1 episode
• Adam — 1× in this paper · explained in 2 episodes
• Admissibility — 1× in this paper · explained in 2 episodes
• Average Fragmentation Rate — 1× in this paper
• Average GPU Utilization — 1× in this paper
• Average Job Completion Time — 1× in this paper
• Deep Reinforcement Learning — 1× in this paper · explained in 2 episodes
• Distribution Shift — 1× in this paper · explained in 5 episodes
• Reinforcement Learning — 1× in this paper · explained in 4 episodes

Related papers

• iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems
• SAMM: Sharded Automated Market Maker
• InferScale: GPU-Native KV Injection for Personalized LLM Serving
• Vigil: Accountable Liveness against Selective Silence
• Steelhead: Interleaving Partially Synchronous and Asynchronous Commit Rules on a Shared DAG
• Pushing CPU Speech Synthesis to the Wall: Extreme Inference Tuning under Serverless Architecture and Billing