Adaptive Virtual Machine Scheduling Algorithms In Cloud And Fog Computing

Due to the large storage capacity and strong computation power of cloud computing,various types of tasks are submitted to cloud data centers and make cloud computing a popular computing model.However,in cloud computing,not only the high computation tasks,but also some delay-sensitive tasks are sent to the remote data centers for processing,which introduces unacceptable latency for the delay-sensitive tasks.Therefore,a computation model called fog computing is proposed.Compared with traditional cloud computing model,it deploys close-to-user equipment to service delay-sensitive applications.Whether cloud computing or fog computing,service providers want to maximize their benefits by virtualization technology and scheduling algorithms.However,existing scheduling algorithms are mainly based on the prior resource requirement assumptions,whose scheduling strategy cannot be adaptively adjusted according to actual usage.Therefore,in this paper,we design reinforcement learning based adaptive virtual machine scheduling algorithms.Compared with the previous works,our main contributions are as follows: 1.We survey the existing works,study the characteristics of cloud and fog computing,and formulate the utility of cloud and fog computing based on their characteristics;2.Different from the previous works,our adaptive algorithms are based on reinforcement learning algorithms instead of bin-packing algorithms and prior distribution assumption based machine learning algorithms.In addition,we optimize our algorithms via deep neural networks to handle large-scale scheduling problems;3.Our experiments are based on real data sets,whose results show that our algorithms can achieve better utility than traditional algorithms.