• Computer Science > Artificial Intelligence [Submitted on 14 Jan 2026 (v1), last revised 20 Feb 2026 (this version, v2)] Title:Cluster Workload Allocation: Semantic Soft Affinity Using Natural Language Processing View PDFAbstract:Cluster workload allocation often requires complex configurations, creating a usability gap. • This paper introduces a semantic, intent-driven scheduling paradigm for cluster systems using Natural Language Processing. • The system employs a Large Language Model (LLM) integrated via a Kubernetes scheduler extender to interpret natural language allocation hint annotations for soft affinity preferences. • A prototype featuring a cluster state cache and an intent analyzer (using AWS Bedrock) was developed. • Empirical evaluation demonstrated high LLM parsing accuracy (>95% Subset Accuracy on an evaluation ground-truth dataset) for top-tier models like Amazon Nova Pro/Premier and Mistral Pixtral Large, significantly outperforming a baseline engine. • Scheduling quality tests across six scenarios showed the prototype achieved superior or equivalent placement compared to standard Kubernetes configurations, particularly excelling in complex and quantitative scenarios and handling conflicting soft preferences.
Article Summaries:
- Researchers have demonstrated that large language models (LLMs) can simplify cluster workload scheduling by interpreting natural‑language “soft affinity” hints. The prototype, built as a Kubernetes scheduler extender, uses an LLM (Amazon Nova Pro/Premier, Mistral Pixtral Large) to parse intent annotations with over 95 % subset accuracy on a ground‑truth dataset. In six benchmark scenarios, the system matched or outperformed standard Kubernetes placement, especially in complex, quantitative cases and when soft preferences conflicted. The study notes synchronous LLM calls introduce latency, recommending asynchronous processing for production use. Overall, the work validates semantic soft affinity as a viable, user‑friendly scheduling approach.
Sources: