[Paper Review] Information-Centric Wireless Networks with Mobile Edge Computing

Yuchen Zhou, F. Richard Yu+2 more|arXiv (Cornell University)|Jun 29, 2017

Caching and Content Delivery17 references3 citations

TL;DR

This paper proposes a virtualized heterogeneous network framework integrating mobile edge computing (MEC) and in-network caching to jointly optimize communication, computing, and caching resources. Using a distributed ADMM-based algorithm, it achieves low-complexity, low-overhead resource allocation that significantly improves total utility, especially with content caching and computation offloading.

ABSTRACT

In order to better accommodate the dramatically increasing demand for data caching and computing services, storage and computation capabilities should be endowed to some of the intermediate nodes within the network. In this paper, we design a novel virtualized heterogeneous networks framework aiming at enabling content caching and computing. With the virtualization of the whole system, the communication, computing and caching resources can be shared among all users associated with different virtual service providers. We formulate the virtual resource allocation strategy as a joint optimization problem, where the gains of not only virtualization but also caching and computing are taken into consideration in the proposed architecture. In addition, a distributed algorithm based on alternating direction method of multipliers is adopted to solve the formulated problem, in order to reduce the computational complexity and signaling overhead. Finally, extensive simulations are presented to show the effectiveness of the proposed scheme under different system parameters.

Motivation & Objective

• Address the growing demand for efficient video delivery and computation in wireless networks by integrating MEC and in-network caching.
• Overcome the limitations of existing systems that treat MEC and caching as separate problems, leading to suboptimal resource utilization.
• Enable dynamic sharing of communication, computing, and caching resources among users from different virtual service providers via network virtualization.
• Formulate a joint optimization problem that captures gains from virtualization, caching, and computing to maximize overall system utility.
• Design a distributed algorithm to reduce computational complexity and signaling overhead while maintaining performance

Proposed method

• Design a virtualized heterogeneous networks (HetNets) framework where mobile virtual network operators (MVNOs) create virtual networks based on service provider (SP) demands.
• Model the system with two types of SPs: one with caching (SP1) and one without (SP2), enabling differentiated service provisioning.
• Formulate a joint optimization problem for virtual resource allocation that maximizes total utility, incorporating gains from virtualization, caching, and computing.
• Use the alternating direction method of multipliers (ADMM) to decompose the coupled problem into subproblems, enabling distributed computation and reducing signaling overhead.
• Introduce primal and dual feasibility conditions with tolerances (υ_pri and υ_dual) to determine convergence of the ADMM algorithm.
• Implement binary recovery based on marginal benefit computation to determine optimal user associations and resource allocations.

Experimental results

Research questions

• RQ1 How can mobile edge computing and in-network caching be jointly optimized in a virtualized wireless network to improve system utility?
• RQ2 What is the impact of integrating virtualization with caching and computing on resource allocation efficiency and system performance?
• RQ3 Can a distributed ADMM-based algorithm achieve near-optimal performance with reduced computational complexity and signaling overhead compared to centralized schemes?
• RQ4 How does the inclusion of caching affect the total utility of the MVNO, especially in scenarios with high reuse of popular video content?
• RQ5 What is the convergence behavior and performance trade-off of the proposed ADMM-based algorithm under varying system parameters?

Key findings

• The proposed ADMM-based distributed algorithm converges rapidly to a stable solution, with convergence speed increasing as the penalty parameter ρ increases.
• The scheme with caching achieves significantly higher total utility than the distributed scheme without caching, especially when popular content is reused.
• The proposed scheme performs close to the centralized optimal solution, despite reduced signaling overhead for content distribution and CSI exchange.
• As the number of users increases, the total utility of all schemes increases, but the proposed scheme maintains a performance advantage due to caching gains.
• Setting ρ = 2 achieves a favorable balance between convergence speed and solution accuracy, making it optimal for the simulation setup.