ARRC: Explainable, Workflow-Integrated Recommender for Sustainable Resource Optimization Across the Edge-Cloud Continuum

2025-07-16

AI Summary

Dynamic multi-tenant edge-cloud collaborative systems face a fundamental trade-off among security, operational transparency, and cost-effective maintenance. Method: This paper proposes a sustainable, low-intrusion resource optimization framework grounded in software engineering principles. It employs an interpretable recommendation architecture wherein optimization logic is encapsulated within auditable agents that coordinate via standardized interfaces. The framework integrates eXplainable AI (XAI), GitOps, cross-layer resource modeling, and workflow embedding to ensure recommendations are traceable, verifiable, and incrementally deployable. Contribution/Results: Its core innovation is the “operations-as-closed-loop” paradigm: recommendations are triggered via ticketing systems, delivered as pull requests, approved by operators, and automatically executed. Evaluated across multi-region industrial deployments, the framework reduces operational effort by over 50%, improves resource utilization by 7.7×, and maintains an error rate below 5%, demonstrating strong engineering feasibility and practical efficacy.

Technology Category

Machine Learning: Learning on the Edge & Model CompressionPlanning, Routing, and Scheduling: Optimization of Spatio-temporal SystemsSearch and Optimization: Distributed Search

Application Category

Systems and Infrastructure for Web, Mobile and WoT: Virtualization and resource management in Web systems and infrastructuresUser Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systemsResponsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Achieving sustainable, explainable, and maintainable automation for resource optimization is a core challenge across the edge-cloud continuum. Persistent overprovisioning and operational complexity often stem from heterogeneous platforms and layered abstractions, while systems lacking explainability and maintainability become fragile, impede safe recovery, and accumulate technical debt. Existing solutions are frequently reactive, limited to single abstraction layers, or require intrusive platform changes, leaving efficiency and maintainability gains unrealized. This paper addresses safe, transparent, and low-effort resource optimization in dynamic, multi-tenant edge-cloud systems, without disrupting operator workflows or increasing technical debt. We introduce ARRC, a recommender system rooted in software engineering design principles, which delivers explainable, cross-layer resource recommendations directly into operator workflows (such as tickets and GitOps pull requests). ARRC encapsulates optimization logic in specialized, auditable agents coordinated via a shared interface, supporting maintainability and extensibility through transparency and the ability to inspect both recommendations and their rationale. Empirical evaluation in a multi-region industrial deployment shows that ARRC reduces operator workload by over 50%, improves compute utilization by up to 7.7x, and maintains error rates below 5%, with most benefits achieved through incremental, operator-approved changes. This demonstrates that explainable, recommendation-based architectures can achieve sustainable efficiency and maintainability improvements at production scale. ARRC provides an empirically evaluated framework for integrating explainable, workflow-driven automation into resource management, intended to advance best practices for robust, maintainable, and transparent edge-cloud continuum platforms.

Problem

• Achieving sustainable resource optimization in edge-cloud continuum
• Reducing operational complexity from heterogeneous platforms
• Providing explainable recommendations without disruptive changes

Innovation

• Explainable recommender system for resource optimization
• Workflow-integrated cross-layer recommendations
• Auditable agents with shared interface