A Cost-Aware Reinforcement Learning Approach for Heterogeneous Multi-Cloud Telecom Operator Infrastructures

• Billy Elly

  Lautech

  Author

Telecommunications operators increasingly deploy workloads across heterogeneous multi-cloud environments to balance cost, latency, and service-level agreement (SLA) compliance. However, traditional threshold-based auto-scaling mechanisms remain dominant in operational deployments despite generating significant cost inefficiencies under dynamic workload conditions . This study addresses the gap between computationally expensive deep reinforcement learning approaches and simplistic reactive scaling by proposing a cost-aware reinforcement learning framework tailored for telecom operator infrastructure. The research employs a design-based methodology combining retrospective analysis of synthetic workload traces calibrated to telecom traffic patterns with prospective simulation using a multi-objective reward function that jointly optimizes operational cost, energy consumption, and SLA satisfaction . The proposed framework achieves a cost reduction of 89.4% relative to static provisioning baselines while maintaining 98.2% SLA compliance under heterogeneous cloud environments . Key findings demonstrate that a balanced reward weighting of 0.4 cost, 0.3 energy, and 0.3 SLA yields the most robust trade-off across workload intensities. The study contributes a replicable reward engineering methodology and establishes that lightweight reinforcement learning agents can bridge the intelligence gap in multi-cloud telecom orchestration without requiring excessive computational overhead. Practical implications include actionable guidance for network operators seeking to reduce total cost of ownership while preserving carrier-grade service reliability.

• pdf

Copyright (c) 2026 Billy Elly (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.