Gap found in cloud edge iot systems for llm based resource control

Smart Adaptive Computing Across the Continuum: LLMs in IoT-Edge-Cloud Resource Management

Distributed, Parallel, and Cluster ComputingArtificial IntelligenceMultiagent Systems

Summary

Managing computing resources that span from tiny devices (IoT) to large cloud servers is tricky because conditions keep changing. The authors look at how two AI tools—deep reinforcement learning (DRL) and large language models (LLMs)—can work together to make smarter management decisions. They extend an existing classification system to better understand how LLMs are used and how feedback is fed back into the system. Their review of recent systems finds a key missing piece: no system fully uses LLM-driven control while getting complete feedback from all parts of the system. They identify this as a challenge needing better communication across different system layers.

What this means in practice

  • For iot system architects: Design orchestration systems that more effectively integrate LLM-driven decisions with complete feedback across IoT, edge, and cloud layers.
  • For cloud platform engineers: Improve multi-tier resource management by developing new feedback abstractions to unify signals for LLM-based orchestration.

A survey. It maps existing work.

Authors

Antonino Vaccarella, Lanpei Li, Vincenzo Lomonaco, Massimo Coppola

Abstract

Managing resources across IoT, edge, and cloud layers calls for continuous, context-aware decisions under constraints that rarely stay fixed. Deep reinforcement learning (DRL) handles this class of problems well, and large language models (LLMs) are increasingly used to augment DRL pipelines, yet the architectural relationship between the two is seldom made explicit. We build on Wang et al.'s taxonomy of Continuum Orchestration Systems employing DRL techniques and extend it with two further dimensions. The AI Augmentation Paradigm measures how LLMs are exploited, while the Feedback channel captures whether and through which system path the execution feedback returns to the LLM in order to close the MAPE control loop at the LLM Orchestration layer. We apply this taxonomy to six recent system architectures and find a common gap, as none combines full LLM orchestration with full agent-layer feedback in a Cloud Continuum setting. We relate this gap to a missing cross-tier feedback abstraction, bridging the incommensurable per-tier signals and the LLM Orchestrator.
View PDFOpen arXiv