CNS Core: Small: DeepEdge: QoE-based Resource Allocation for Future Heterogeneous and Dynamic Edge-IoT Applications

点击翻译按钮获取中文摘要

The emerging Internet of Things (IoT) is connecting increasing numbers of smart devices and enabling varieties of heterogeneous IoT applications, empowered by cloud and edge computing technologies. In particular, edge or fog computing technologies will significantly benefit IoT applications that are delay-sensitive, bandwidth/data intensive, or that require closer intelligence. However, for effective Edge-IoT resource allocation, significant challenges exist due to the following requirements and constraints: 1) the demand side that a massive number of IoT devices can run heterogeneous applications with various Quality of Service (QoS) requirements and different priorities; and 2) the supply side that the edge clouds need to dynamically and optimally allocate limited and multidimensional resources (CPU, storage, and bandwidth) at geospatially distributed points. The objective of this project is to respond to these challenges by designing and developing a new Edge-IoT framework named DeepEdge using deep online learning that allocates resources to heterogeneous IoT applications and dynamic IoT devices to maximize users' Quality of Experience (QoE). The proposed research will advance knowledge and fundamentally change the way future edge computing systems work in supporting heterogeneous and dynamic IoT applications. The transformative research outcomes will benefit users and society with inexpensive and effective IoT application delivery, and contribute to important societal challenges in supporting emerging IoT devices and applications. The project will also broadly involve and impact K-12 underrepresented groups and female students in computer science, and develop strong research and education integration for various levels of students. The goal of the proposed research is to develop the framework, model, and algorithms in effectively delivering heterogeneous IoT applications on edge clouds and provisioning high-quality QoE for users. More specifically, the project will result in: i) a new QoE model to quantify the users satisfaction and its related factors including multiple applications QoS requirements and applications priority; ii) a new deep machine learning based two-stage resource allocation scheme that will adapt application QoS requirements according to available resources at the edge cloud and maintain application's priority, and will intelligently and jointly allocate communication and computation resources with the objective of maximization of users QoE; iii) a novel deep Q-learning scheme that will dynamically select the most appropriate edge nodes to handle the multiple application tasks with the goal to optimize the task execution delay; iv) a hardware and software test-bed implementation to validate and evaluate the effectiveness, efficiency, and the practicality of the proposed research.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.