QoE Based Revenue Maximizing Dynamic Resource Allocation and Pricing for Fog-Enabled Mission-Critical IoT Applications
QoE Based Revenue Maximizing Dynamic Resource Allocation and Pricing for Fog-Enabled Mission-Critical IoT Applications
复制标题
DOI：
10.1109/tmc.2020.2999895
复制
发表时间：
2020-06
影响因子：
7.9
通讯作者：
Muhammad Junaid Farooq;Quanyan Zhu
中科院分区：
文献类型：
--
作者：
Muhammad Junaid Farooq;Quanyan Zhu
雾计算正在成为物联网（IoT）应用的重要组成部分，充当其计算引擎。关键任务物联网应用程序对延迟高度敏感，这取决于云服务器的物理位置。云服务提供商（CSP）可以使用不同响应速率的雾节点，它面临着将顺序接收到的物联网数据转发到其中一个雾节点进行处理的挑战。由于请求的到达时间和性质是随机的，因此必须实时对请求进行最佳分类，并在雾节点上分配可用的虚拟机实例（vmi），以便为用户提供高QoE，从而为CSP产生更高的收入。本文采用基于应用分配结果QoE的定价策略，根据计算请求的统计信息得到最优动态分配规则。与文献中的其他静态匹配方案相比，开发的解决方案在统计上是最优的，动态的，并且可以实时实现。通过仿真对该框架的性能进行了评估，结果表明与基准方案相比，该框架有了显著的改进。
Fog computing is becoming a vital component for Internet of things (IoT) applications, acting as its computational engine. Mission-critical IoT applications are highly sensitive to latency, which depends on the physical location of the cloud server. Fog nodes of varying response rates are available to the cloud service provider (CSP) and it is faced with a challenge of forwarding the sequentially received IoT data to one of the fog nodes for processing. Since the arrival times and nature of requests is random, it is important to optimally classify the requests in real-time and allocate available virtual machine instances (VMIs) at the fog nodes to provide a high QoE to the users and consequently generate higher revenues for the CSP. In this paper, we use a pricing policy based on the QoE of the applications as a result of the allocation and obtain an optimal dynamic allocation rule based on the statistical information of the computational requests. The developed solution is statistically optimal, dynamic, and implementable in real-time as opposed to other static matching schemes in the literature. The performance of the proposed framework has been evaluated using simulations and the results show significant improvement as compared with benchmark schemes.