As a newly emerging computing paradigm, mobile edge computing (MEC) can energize novel mobile applications, especially the ultra-latency-sensitive ones, by providing powerful local computing capabilities and lower end-to-end delays. However, considering the complex cyber-physical environment and varied time-space user behaviors, as mobile application vendors, how to decide which edge server to serve which edge user in real-time becomes the key challenge. Instead of assuming the simultaneous-batch-arrival pattern of incoming edge users and handling the edge user allocation (EUA) problem as a static optimization, in this paper, we consider the online EUA problem where edge users’ resource demands arrive and depart dynamically. We consider the long-term edge user allocation rate, edge server hiring cost, and edge server energy consumption as allocation targets from the mobile application vendor perspective, and propose a decentralized reactive approach by employing a fuzzy control mechanism to yield the real-time allocation decisions. Experiments based on real-world MEC environment datasets demonstrate our approach outperforms state-of-the-art and baseline ones in terms of user allocation rate, server hiring cost, and energy consumption.

作为一种新兴的计算范例，移动边缘计算（MEC）通过提供强大的本地计算功能和较低的端到端延迟，可以激发新颖的移动应用程序，尤其是对超时延敏感的应用程序。但是，考虑到复杂的网络物理环境和不同的时空用户行为，作为移动应用程序供应商，如何决定实时为哪个边缘用户服务的边缘服务器成为关键挑战。在本文中，我们没有考虑传入的边缘用户的同时到达批次模式并将边缘用户分配（EUA）问题作为静态优化来处理，而是考虑了在线EUA问题，其中边缘用户的资源需求到达并动态地离开。我们考虑了长期边缘用户分配率，边缘服务器雇用成本，从移动应用程序供应商的角度来看，将边缘服务器能耗和边缘服务器能耗作为分配目标，并通过采用模糊控制机制来提出实时分配决策，提出了一种分散式无功方法。基于实际MEC环境数据集的实验表明，在用户分配率，服务器租赁成本和能耗方面，我们的方法优于最新方法和基准方法。