Modeling and Distributed Optimization for Cloud/Fog Computing Networks

Speak:     Prof. Yong Xiao
Time:       09:00-10:00, Nov. 14
Location:  SIST 1C 502
Host:       Prof. Yang Yang
Abstract:
Fog computing is a virtualized network architecture that uses one or a collaborative multitude of end-user clients or near-user edge devices to carry out a substantial amount of storage, communication, and control, configuration, measurement, and management. It can offload workload from cloud data centers, reduce the transmission latency, improve the system reliability, and ease traffic congestion of the Internet. It also enables many new services and applications that cannot fit well in the traditional cloud computing architecture. In this talk, I will first briefly introduce the recent developments in the mobile cloud and fog computing. I will then introduce two important performance metrics for fog computing networks: quality-of-experience (QoE) and power efficiency. A fundamental tradeoff between QoE and power efficiency will then be discussed. Motivated by the observation that the users QoE can be further improved if the workload offloading process of each fog node can be helped by others, I also talk about a fog computing framework with fog node cooperation. The QoE and power efficiency tradeoff under cooperative fog computing will be discussed. Recently, standardization bodies including 3GPP and ETSI have included fog computing as the key component in 5G networks. We introduce a novel concept called dynamic network slicing for implementing fog computing into the 5Gs service-base architecture (SBA). In this concept, the limited computational resources can be sliced and reserved according to the traffic demands and Quality-of-Service (QoS) requirements of various supported services. We propose a stochastic overlapping coalition-formation game-based framework to investigate distributed cooperation and joint network slicing between fog nodes under randomly fluctuating resource availability, workload arrival processes, and QoS demands. Applications of 5G-enabled fog computing networks in the Tactile Internet and smart vehicular systems will be discussed. Finally, I will also talk about my current works on designing fog computing-enabled self-driving vehicular systems and deep learning-based optimization algorithms.