AtumAI: A Principled Framework for Agentic Generation of Datacenter Control-Plane Policies

2026-08-03

AI Summary

This work addresses the challenges in data center control plane policy design—namely rapid hardware and software evolution, a vast and highly coupled design space, and development cycles spanning several months. To overcome these limitations, the authors propose AtumAI, a framework that employs a task compiler to automatically translate natural language objectives into machine-verifiable formal specifications. AtumAI integrates diffusion models, evolutionary algorithms, and surrogate models within a cooperative search loop to systematically generate, evaluate, and optimize policies. This approach enables, for the first time, standardized, transferable, and efficient policy exploration: it supports formal constraints, facilitates cross-task knowledge reuse, and transcends the candidate limitations inherent in large language models. Evaluated on workload placement, resource scaling, and power management tasks, the automatically generated policies significantly outperform expert-designed solutions, reducing new task integration from months to merely specifying the objective.

Abstract

The efficiency of a datacenter rests on its control plane policies. Designing these policies is increasingly hard: the hardware-software stack grows fast, the design space is vast and interdependent, and prototyping a single policy takes months. Agentic AI promises to automate this search. Off the shelf, however, it falls short on three fronts. It is not formal: with no structured, searchable statement of the problem, the search has little structure to exploit and hard constraints are not guaranteed. It is not transferable: each task is solved from scratch, so nothing learned on one task carries to the next. Finally, it is not systematic: relying on the LLM as the sole source of candidates, it explores a narrow slice of the design space and settles into local optima. We introduce AtumAI, a framework that generates datacenter control-plane policies with agentic AI, making the process formal, transferable, and systematic. From a goal stated in plain language, AtumAI autonomously proposes, tests, and refines candidate policies until one satisfies the request. It does so through two components. The Datacenter Task Compiler automates problem formulation: it compiles the request into a formal, machine-checkable, and searchable specification of the task's objectives, constraints, decision variables, and evaluation methodology. The Evolutionary Design Discovery Loop then searches this specification, expanding the search beyond the LLM itself via a diffusion model, an evolutionary algorithm, and a surrogate model. Together, they reduce onboarding a new task from months of engineering to writing its description. We evaluate AtumAI on three control-plane tasks with distinct problem scopes, design spaces, and trade-offs: workload placement, resource scaling, and power management. Across all tasks, the policies generated by AtumAI consistently outperform expert-engineered baselines.

Problem

• datacenter control-plane policies
• agentic AI
• policy design
• formal specification
• design space exploration

Innovation

• Agentic AI
• Formal Specification
• Evolutionary Design
• Control-Plane Policy
• Transferable Optimization