gEdge:A Container-Based Cloud-Edge Collaboration Framework for Heterogeneous Computing
The advantages of flexible on-demand provisioning,high availability,and high resource utilization have made cloud computing technology the dominant computing paradigm of the past decade.With the advent of the Internet of Everything era,relying on cloud computing technology alone can no longer meet the demands of hundreds of millions of IoT(Internet of Things)devices and their data traffic.Edge computing can be seen as an evolution of cloud computing,emerging from the rise of 5G networks and the IoT.With the widespread use of cloud gaming,VR(Virtual Reality)technology,and artificial intelligence technology in daily life,the demand for computing resources is growing day by day.Restricted by size,weight,and power,the node devices at the edge have weak computing resources.In this paper,we propose gEdge,a cloud-edge collaboration framework for heterogeneous computing based on container technology.The framework divides the physical GPU resources in the cloud into multiple virtual GPU resources,provides GPU compute resources for the edge nodes on demand,and is transparent to user containers,through GPU virtualization technology.Experiments show that the use of the gEdge framework can reduce the container image size used by edge nodes by 48.8％,container start-up time by 35.5％,and average relative running speed by 213％.
万方数据
维普
