- The paper demonstrates that multi-user computation offloading is NP-hard and introduces a game-theoretic framework that ensures a Nash equilibrium.
- It proposes a distributed algorithm operating in slotted time that converges quickly using best response strategies under multi-channel interference.
- Numerical results indicate up to 30% improvement in beneficial cloud computing users and a significant reduction in system-wide computation overhead.
Efficient Multi-User Computation Offloading for Mobile-Edge Cloud Computing
Abstract:
The paper explores the multi-user computation offloading problem within the context of mobile-edge cloud computing under a multi-channel wireless interference environment. The authors demonstrate the NP-hard nature of computing a centralized optimal solution and propose a game-theoretic framework for efficient, distributed computation offloading. They formulate this as a multi-user computation offloading game, prove the existence of a Nash equilibrium, and establish its finite improvement property. Moreover, a distributed computation offloading algorithm is proposed, and its performance is evaluated through numerical results.
Introduction:
The proliferation of resource-intensive mobile applications has introduced significant challenges in managing the constrained computation resources and battery life of mobile devices. Mobile cloud computing has emerged as a viable solution, allowing offloading of computational tasks to external clouds. However, traditional approaches that rely on remote public clouds suffer from high latency issues. The concept of mobile-edge cloud computing, which places cloud capabilities at the edge of radio access networks close to mobile users, offers an alternative by providing lower latency connections. Nonetheless, efficient coordination of wireless access among multiple mobile users remains a critical challenge.
System Model:
The system model defines the environment with a set of mobile device users and a base-station supporting multi-channel wireless communication. The model incorporates both communication and computation aspects, with detailed mathematical formulations for the rates and overheads of local and offloaded computing. The critical factor of interference among users sharing the same wireless channel is studied, establishing thresholds for beneficial cloud computing, where offloading to the cloud does not incur higher overhead than local computation.
Game-Theoretic Framework:
Due to the NP-hard nature of the centralized optimization problem, the authors propose a game-theoretic approach. They model the computation offloading problem as a strategic game where each user aims to minimize its computation overhead by choosing between local and cloud computing across multiple channels. A Nash equilibrium is established, where no user can further reduce its overhead by unilaterally changing its strategy. The potential game framework ensures that the game admits a Nash equilibrium and possesses the finite improvement property.
Proposed Algorithm:
A distributed computation offloading algorithm is designed to achieve the Nash equilibrium. The algorithm operates in a slotted time structure, where users measure interference in each decision slot and then update their computation offloading decisions based on their best response strategies. The analysis reveals that the algorithm converges within a bounded number of slots and scales efficiently with the number of users.
Performance Analysis:
The performance of the proposed algorithm is evaluated in terms of two metrics: the number of beneficial cloud computing users and the system-wide computation overhead. The price of anarchy (PoA) is used to quantify the efficiency ratio of the worst-case Nash equilibrium over the centralized optimal solutions. The authors provide bounds on the PoA, demonstrating that the distributed computation offloading algorithm's performance is robust and close to the centralized optimal under varying conditions.
Extension to Wireless Contention Model:
The authors extend their paper to a scenario where the wireless channels are accessed in a contention-based manner rather than interference-based. They show that the multi-user computation offloading game retains its potential game properties under this model as well. The same distributed algorithm can be applied, achieving similar performance guarantees.
Numerical Results:
Extensive numerical results corroborate the theoretical findings, demonstrating the algorithm's efficiency and scalability. The proposed distributed algorithm outperforms naive approaches such as all-local or all-cloud computing in various scenarios. It achieves up to 30% performance improvement in the number of beneficial cloud computing users and significant reductions in system-wide computation overhead.
Conclusion:
The paper presents a rigorous and comprehensive paper of the multi-user computation offloading problem in mobile-edge cloud computing. The game-theoretic approach provides a robust framework, and the proposed distributed algorithm offers practical scalability and performance benefits. Future work will explore dynamic user mobility and joint power control within the offloading paradigm.