A Kubernetes-Oriented Edge Network Orchestrator for Heterogeneous Environment
Chenyu Lu, Zhaowu Huang, Caishan Weng, Feng Jiao, Xiaolin Guo, Fang Dong · 2022
Edge computing provides computing resources for devices at the network edge side close to the data source, so as to promote the Quality of Services (QoS) of applications that provide to end devices. In recent years, Kubernetes (k8s) is widely concerned because of the advantages that can easily integrate edge resources and realize edge computing systems. However, due to k8s only having simple dispatching algorithms like first-come-first-serve, it cannot achieve efficient request dispatching in a complex heterogeneous environment with limited resources. In this paper, we design and implement a k8s-oriented edge network orchestrator, aiming at minimizing the completion time of all devices’ tasks in an edge environment. At the theoretical level, we model the joint problem of resource allocation and task dispatching as a mixed integer programming problem. We first determine the optimal resource allocation policy with a given task offloading decision profile. Then, a decentralized iterative improve dispatching (IID) algorithm is proposed to efficiently calculate the task dispatching decisions. At the system level, we implement a prototype k8s-based edge system that equips the network orchestrator which embeds with the IID algorithm. The orchestrator provides the task-dispatching service through docker containers. Experimental evaluation results verify that the orchestrator has a significant progress compared with the state-of-the-art methods, achieving up 2.21 times on average.