Video Analysis Task Scheduling In The Edge Environment

With the development of Internet of Things,more and more front-end devices are deployed in the network edge continuously collecting useful information.However,compared to the ever increasing data size,the wide area network bandwidth has come to a standstill,thus cloud-based data processing may incur excessive transmission delay.Edge computing is a new computing paradigm which advocates processing data at the edge of the network.Since edge servers are placed in proximity to users,the problem of high network delay can be mitigated.Emerging applications such as cognitive assistance,mobile gaming,virtual reality and augmented reality all rely on effective analysis of videos from mobile devices in real time.Leveraging edge computing,these applications can offload computationintensive tasks to nearby edge servers.In this paper we focus on the task scheduling for various video analysis applications under different circumstances in the edge environment.The main contributions of this paper are summarized as follows.(1)We consider the offloading of MAR applications comprising multiple tasks,over a generic cloud-edge computing system including a group of heterogeneous edge servers and remote clouds.In this paper,a MAR application is modeled as a chain of inter-dependent tasks with data transmissions between them,each task can be offlfloaded to different servers.We design both offline and online algorithms by optimizing the server assignment and the resource allocation jointly.The experimental results show that both our offline and online algorithms can significantly reduce the service delays of MAR applications when compared to other heuristics.(2)we study the configuration adaption and bandwidth allocation for multiple video streams,which are connected to the same edge node sharing an upload link.Wepropose an efficient online algorithm,called JCAB,which jointly optimizes configuration adaption and bandwidth allocation to address a number of key challenges in edgebased video analytics systems,including edge capacity limitation,unknown network variation,intrusive dynamics of video contents.Our algorithm is developed based on Lyapunov optimization and Markov approximation,works online without requiring future information,and achieves a provable performance bound.Simulation results show that JCAB can effectively balance the analytics accuracy and energy consumption while keeping low system latency.