An Optimal Deployment Framework for Multi-Cloud Virtualized Radio Access Networks

Abstract

Virtualized radio access networks (vRAN) are emerging as a key component of wireless cellular networks, and it is therefore imperative to optimize their architecture. vRANs are decentralized systems where the Base Station (BS) functions can be split between the edge Distributed Units (DUs) and Cloud computing Units (CUs); hence they have many degrees of design freedom. We propose a framework for optimizing the number and location of CUs, the function split for each BS, and the association and routing for each DU-CU pair. We combine a linearization technique with a cutting-planes method to expedite the exact problem solution. The goal is to minimize the network costs and balance them with the criterion of centralization, i.e., the number of functions placed at CUs. Using data-driven simulations we find that multi-CU vRANs achieve cost savings up to 28% and improve centralization by 77%, compared to single-CU vRANs. Interestingly, we see non-trivial trade-offs among centralization and cost, which can be aligned or conflicting based on the traffic and network parameters. Our work sheds light on the vRAN design problem from a new angle, highlights the importance of deploying multiple CUs, and offers a rigorous optimization tool for balancing costs and performance.

Deep Reinforcement Based Optimization of Function Splitting in Virtualized Radio Access Networks

“…Table I describes the particular vRAN split options and their requirements. Our model refers to the standardization of 3GPP [1], [2] and seminal white paper [3], where each split has a different performance gain [6], [11]. Split 0: All functions are at DU, except the RF layer is at RU.…”

“…Going from Split 1 to 3, more functions are hosted at CU. In addition to increasing network performance, a higher centralization level can lead to more cost-saving [11]. However, centralizing more functions increases the data load to be transferred to CU, going from λ in S0 to 2.5 Gbps in S3 for each BS, and has stricter delay requirements (Table I).…”

“…Follow up works, [8] and [9] offered optimal solution of minimizing total cost for integration vRAN with Mobile Edge Computing (MEC). Then, [10] proposed an optimized multi-cloud vRAN framework with balancing its centralization [11]. However, the mentioned works above rely on mathematical optimization techniques that often have a slow convergence rate and exponential complexity for finding the optimal solution, particularly in large networks.…”

“…It is worth noting that our approach requires minimal handcrafted engineering. It does not need to know the vRAN split problem mathematically, e.g., mathematical optimization-based approaches [8]- [11], or direct access to the optimal labeled data, e.g., supervised learning [21]. Instead, it learns from interaction with the environment that expects to receive the reward (total network cost) signal and Penalization (constraints violation).…”

“…We evaluate our approach in a synthetic network generated by the Waxman algorithm that highly represents a backhaul network [26]. The used system parameters are from a measurement-based 3GPP-compliant system model [10], [11]. To assess our approach's effectiveness, we compare it to the optimal value obtained from a Phyton-MIP solver 1 .…”