Network function virtualization (NFV) implements mobile edge computing (MEC) services as software appliances, and allows resources to be adaptively allocated to accommodate demand variations. Scalability and network cost (including operational cost and response latency) are key challenges. This paper presents a new game-theoretic approach to minimizing the network cost, where access points (APs) select MEC servers and routes in a decentralized manner, and unloaded routers and links are deactivated for cost saving. The key idea is that we interpret the minimization of network cost as a mixed game with a non-monotonic cost function capturing both the operational cost and response latency. We prove that the game is conditionally an ordinary potential game and converges to α\alpha -approximate equilibriums. A closed-form expression is derived for the convergence delay. Another important aspect is that we integrate Stackelberg routing into the proposed mixed game to avoid inefficient equilibriums (with high cost or latency). We prove that the mixed game can converge faster to better equilibriums under linear response latency models. Extensive simulations corroborate the new game-theoretic approach can significantly outperform existing techniques in terms of efficiency, convergence, and scalability.

中文翻译：

---

用于移动边缘计算的去中心化路径选择和睡眠调度的新博弈论方法


网络功能虚拟化 (NFV) 将移动边缘计算 (MEC) 服务实现为软件设备，并允许自适应分配资源以适应需求变化。可扩展性和网络成本（包括运营成本和响应延迟）是主要挑战。本文提出了一种新的博弈论方法来最小化网络成本，其中接入点（AP）以分散的方式选择MEC服务器和路由，并停用卸载的路由器和链路以节省成本。关键思想是，我们将网络成本最小化解释为混合游戏，其中非单调成本函数捕获运营成本和响应延迟。我们证明该博弈有条件地是一个普通的潜在博弈，并且收敛于 α\alpha 近似均衡。导出了收敛延迟的封闭式表达式。另一个重要的方面是我们将 Stackelberg 路由集成到提议的混合游戏中，以避免低效均衡（具有高成本或延迟）。我们证明混合博弈可以在线性响应延迟模型下更快地收敛到更好的平衡。广泛的模拟证实了新的博弈论方法在效率、收敛性和可扩展性方面可以显着优于现有技术。