Data Replication and Placement in the Cloud–Edge–Terminal Framework: A Systematic Review of AI-Driven, Multi-Objective Strategies

Abstract

Efficient data replication approaches are essential to address the critical latency-cost trade-offs due to rapid growth in data-intensive applications. Centralized solutions based on traditional cloud models often suffer from high transmission delays, while static heuristic models struggle to manage multi-dimensional uncertainty in decentralized environments. This systematic literature review (SLR) offers a comprehensive synthesis of both advanced and emerging data replication and placement techniques in the Cloud–Edge–Terminal Framework. Based on the PRISMA guidelines, 62 primary studies were included from 1096 initial records identified by screening five large-scale resources from 2019 to June 2026. The results show a shift from cloud-centric infrastructures to known adaptations of adaptive and AI-driven orchestrations in cloud-edge-terminal environments, such as low-earth orbit satellite systems and quantum-safe replication. An important trend is the move to deep reinforcement learning and pre-processed geometric oracles, increasing execution speed from 101 to 108 compared to classical methods. Optimizing these approaches increasingly targets multi-objective trade-offs among performance quality, cost, energy efficiency, and reliability. This paper summarizes a unified technical roadmap highlighting four main open challenges: i.e., serverless cold-start bottlenecks, mobility in LEO satellite deployments, lack of multi-vendor validation in empirical studies, and security weaknesses in zero-trust architectures. Future research should focus on federated learning and edge-aware protocols to support a resilient and secure global computing ecosystem.