Abstract
As an emerging paradigm enabling mobile devices to leverage additional computation resources from nearby MEC servers (MSs), mobile edge computing (MEC) has drawn great attention from academia to industry. Unlike the conventional cloud server, the MEC provides a medium-scale and portable computation ability at MSs without relying on the time-consuming and capacity-constrained backhaul. However, the MEC offloading process is still highly sensitive to the fluctuation of both radio and computing resources. In this paper, considering the independent variation of the wireless channel conditions and computing tasks, we propose a Mixed-timescale Joint Computational offloading and Wireless resource allocation (MJCW) algorithm for latency-critical applications, aiming at minimizing the total energy consumption. Through such a new approach, the original NP-hard problem is decoupled into a short-term stage problem seeking for the allocation of physical power and subcarrier and a long-term stage problem of task offloading and frequency scaling. The simulation results show that the proposed algorithm achieves excellent performance in energy saving in comparison with conventional schemes and realizes higher utilization of green energy by adjusting the energy price of MSs.
Introduction
With the advent of the 5G and Internet of Things (IoT) era, mobile devices are increasingly required to support resource-intensive applications such as Augmented Reality (AR), Virtual Reality (VR), and high-fidelity mobile gaming. These applications often exceed the local processing capabilities of mobile hardware, leading to significant latency and rapid battery depletion. Mobile Edge Computing (MEC) addresses these challenges by offloading computational tasks to servers located at the network edge.
However, two primary challenges persist in optimizing these systems. First, there is a significant timescale mismatch: wireless channel conditions fluctuate in milliseconds (short-term), whereas task characteristics and server CPU adjustments occur over longer intervals (long-term). Treating these as synchronous often leads to resource inefficiency. Second, future networks must manage energy heterogeneity, integrating traditional grid-powered servers with "Green" MEC servers (GMS) powered by renewable energy. Efficiently prioritizing green energy without compromising performance remains a critical hurdle.
The MJCW Framework
To address these issues, we propose the Mixed-timescale Joint Computational offloading and Wireless resource allocation (MJCW) algorithm. This strategy tackles the complex NP-hard optimization problem by decoupling it into two distinct stages based on their respective timescales.
Short-Term Stage (STOA)
This stage focuses on the physical layer, adaptively allocating subcarriers and transmission power based on instantaneous channel state information (CSI). By employing Lagrangian relaxation and a subgradient algorithm, the system can efficiently solve this non-convex problem in real-time as signal conditions change.
Long-Term Stage
The long-term stage manages the logic of task offloading and server frequency scaling. It determines which server should receive a specific task and utilizes Dynamic Voltage and Frequency Scaling (DVFS) to adjust the server's CPU frequency. To prevent server congestion, the Hungarian algorithm is utilized to establish an optimal one-to-one matching between users and available MEC servers.
Green Energy Integration
A key innovation of the MJCW algorithm is the introduction of an energy price factor ($\mu_k$). This economic steering mechanism allows network operators to prioritize renewable energy by assigning a lower price to green servers. The optimization logic naturally gravitates toward these lower-cost resources, facilitating the transition to sustainable 5G infrastructure without requiring a complete redesign of the offloading logic.
Performance Evaluation
The algorithm was evaluated using a Monte Carlo simulation environment with mobile users and servers randomly deployed within an 80-meter radius. The channel model incorporated Rayleigh fading for small-scale fluctuations and standard path-loss models for large-scale fading, aligned with 5G specifications.
Key Findings
- Energy Efficiency: The MJCW algorithm consistently outperformed baseline strategies, such as the Average Allocation Strategy (AAS) and standalone short-term optimization (STOA), demonstrating the value of the mixed-timescale approach.
- Latency-Energy Tradeoff: Stricter latency requirements (e.g., 40ms) significantly increased energy consumption, as servers were forced to "overclock" CPUs to meet deadlines.
- MEC Diversity: The presence of multiple servers allows the system to "smooth out" wireless randomness; if one link is poor, the system can dynamically select a more favorable server.
- Sustainability: By adjusting the price factor for GMS, the algorithm successfully shifted the majority of the computational load to renewable resources.
Conclusion and Future Work
The MJCW algorithm provides a robust framework for energy-efficient, green MEC systems by effectively decoupling wireless and computational resource management. While the current model assumes single-task execution per server and stationary users, future research will explore distributed control versions, high-speed mobility scenarios (e.g., vehicular networks), and the energy costs associated with data-heavy return paths.
Summary updated