A Tensor-Based Big-Data-Driven Routing Recommendation Approach for Heterogeneous Networks

TLDR

Telecommunication networks are shifting to data‑center‑based architectures that integrate physical, virtual, and orchestration functions, yet heterogeneous factors such as bandwidth, delay, and protocols create significant routing challenges, prompting the use of big‑data techniques. This work introduces a tensor‑based, big‑data‑driven routing recommendation framework spanning edge, fog, cloud, and application planes. The framework employs a hierarchical tensor decomposition to generate efficient routing paths and a tensor‑matching scheme using controlling, seed, and orchestration tensors to realize recommendations. A case study demonstrates the framework’s processing steps and validates its routing recommendation capabilities.

Abstract

Telecommunication networks are evolving toward a data-center-based architecture, which includes physical network functions, virtual network functions, as well as various types of management and orchestration systems. The primary purpose of this type of heterogeneous network is to provide efficient and convenient communication services for users. However, the diverse factors of a heterogeneous network such as bandwidth, delay, and communication protocol, bring great challenges for routing recommendations. In addition, the growing volume of big data and the explosive deployment of heterogeneous networks have started a new era of applying big data technologies to implement routing recommendations. In this article, a tensor-based big-data-driven routing recommendation framework, including the edge plane, fog plane, cloud plane, and application plane, is proposed. In this framework, a tensor-based, holistic, hierarchical approach is introduced to generate efficient routing paths using tensor decomposition methods. Also, a tensor matching method including the controlling tensor, seed tensor, and orchestration tensor is employed to realize routing recommendation. Finally, a case study is used to demonstrate the key processing procedures of the proposed framework.