Mobile edge networks can provide ultra-low latency by deploying services at the edge of networks. However, it is impracticable to place services on all the edge servers due to limited deployment cost requirement. It is desirable to make the optimal edge application placement decisions in a minimum response time and deployment cost, which involve two possibly conflicting objectives. An important issue here is that the number of edge application request from mobile users, which is the key factor determining the realization of two objectives, can vary considerably and the number of edge application request usually unknown before deploying edge application in a particular edge site. It has been found recently that human mobility (location and time features) have a strong influence on what kinds of application that mobile users choose to use. Based on the above observation, we investigate the mobility aware edge service placement problem aiming at optimizing service latency and deployment cost. The problem is formulated as a multi-objective optimization problem and can be solved by multi-objective context multi-armed bandit with a dominant objective. The features of user mobility (time, location) are considered as the context information guiding the edge application placement decisions. We develop mobility-aware edge service placement (MAESP) method and analyse performance measures of MAESP using the 2-dimensional (2D) regret. We show that the 2D regret of MAESP are sublinear in the number of rounds. Based on a real-world dataset, we carry out extensive simulations to evaluate the performance of MAESP. The results show that MAESP outperforms the benchmark algorithms.

通过在网络边缘部署服务，移动边缘网络可以提供超低延迟。但是，由于有限的部署成本要求，无法在所有边缘服务器上放置服务。希望以最小的响应时间和部署成本来做出最佳的边缘应用程序放置决策，这涉及两个可能相互冲突的目标。这里的一个重要问题是，来自移动用户的边缘应用程序请求的数量（决定两个目标实现的关键因素）可能会发生很大变化，并且在将边缘应用程序部署到特定边缘站点之前，边缘应用程序请求的数量通常是未知的。最近发现，人类移动性（位置和时间特征）对移动用户选择使用哪种应用程序有很大的影响。基于以上观察，我们研究了旨在优化服务延迟和部署成本的移动感知边缘服务放置问题。该问题被表述为多目标优化问题，可以通过具有主要目标的多目标上下文多臂匪解决。用户移动性（时间，位置）的特征被视为指导边缘应用程序放置决策的上下文信息。我们开发了可感知移动性的边缘服务放置（MAESP）方法，并使用二维（2D）遗憾分析了MAESP的性能指标。我们证明了MAESP的2D后悔在回合数上是次线性的。基于现实世界的数据集，我们进行了广泛的仿真以评估MAESP的性能。结果表明，MAESP优于基准算法。