Deadline-Driven Hierarchical Agentic Resource Sharing for AI Services and RAN Functions in AI-RAN

TL;DR

This paper introduces a hierarchical agentic framework (HAF) for efficient compute sharing in AI-RAN, combining slow and fast-timescale scheduling with predictive migration filtering, significantly improving service fulfillment and SLO adherence.

Contribution

The novel HAF framework integrates LLM-based slow-timescale placement with a convex algorithm for fast scheduling, enhanced by a predictive critic to minimize service interruptions.

Findings

HAF achieves 90.0% SLO fulfillment, a 20.5% improvement over baselines.

Service request fulfillment increases from 51% to 85.3%.

HAF maintains performance under diverse load conditions.

Abstract

AI-RAN consolidates AI services and Radio Access Network (RAN) functions onto a unified, GPU-accelerated infrastructure at the network edge. However, compute sharing between real-time RAN functions and highly heterogeneous AI services requires coordination of scheduling decisions at mismatched timescales, and placement adaptation may require service migration across nodes with non-negligible interruptions. This paper proposes a hierarchical agentic framework (HAF) for compute sharing in AI-RAN that combines a large language model (LLM)-based agent for slow-timescale placement of AI services and RAN functions with a closed-form, deadline-aware convex algorithm for fast-timescale GPU/CPU allocation. The LLM agent is further equipped with a predictive critic that filters out migrations when the induced service interruption outweighs the expected service-level objective (SLO) benefit. Experimental results show that HAF reaches 90.0% overall SLO fulfillment, a 20.5% improvement over the strongest baseline, and raises AI service request fulfillment from 51% to 85.3%. Further evaluations show that HAF retains its advantage under diverse load conditions, while the critic consistently improves SLO fulfillment across multiple open-source LLM agents.