Guest post by KubeEdge Maintainers

Introduction

KubeEdge is an open-source edge computing platform. Based on the native container orchestration and scheduling capabilities of Kubernetes, KubeEdge achieves functionalities such as cloud-edge synergy, edge computing, edge device management, and edge autonomy. KubeEdge is also widely implemented in scenarios such as 5G MEC and AI cloud-edge synergy by using plug-ins.

This article talks about HarmonyCloud’s edge computing practices in edge-cloud synergy for calling centers and data collection on a cross-sea bridge.

Edge Computing Ecosystem Development

Gartner, a leading global information technology research and advisory firm, forecasts the top 10 strategic technology trends every year. Edge computing has been listed by Gartner as one of the top 10 strategic technology trends for four consecutive years since 2018.

As one of the three mainstream frameworks of the open-source community, KubeEdge is the first cloud-native edge computing framework designed for cloud-edge synergy. Based on the native container orchestration and scheduling capabilities of Kubernetes, KubeEdge enables applications, resources, data, and devices across edges and clouds to collaborate with each other in edge computing.

Edge Computing Scenarios

How to Implement Edge Computing?

Case 1: Cloud-edge synergy platform of China Mobile Online

This is a typical high rate scenario focusing on the cloud and edge, which has been listed in the best practices of distributed cloud and cloud-edge synergy released by CAICT in 2020.

Background

– Been cooperating for three years.

– Jointly built the largest container-based call platform in the world.

– Containerized all services and deployed them in Kubernetes clusters. (2018-2019)

– Developed edge-cloud synergy. (2019-2020)

China Mobile Online built a central cloud in 2018-2019. However, there were 31 branches of the 10086 hotline, and each branch had more than 10 or even hundreds of machines. The problem was that these branches also needed containerization, so that their applications could be released and upgraded in a unified manner by the central cloud of the headquarters. In this regard, compute capabilities were extended from the central cloud to the equipment rooms of each branch to process data while allowing branches to use the same CI/CD process. 

To do so, one approach was to build a Kubernetes cluster in each branch and incorporate them into the central cloud. The other was to treat each machine as a node and add all nodes to the central cloud for unified management.

The first approach may result in low resource utilization as three machines will be occupied as master nodes to build an HA cluster for managing all other machines of a branch. These three nodes cannot process business services anymore. Besides, if multiple clusters are created, O&M will be difficult even K3s, instead of K8s is used. At the end of 2018, China Mobile Online started to explore KubeEdge, and finally chose the second approach.

Edge-Cloud Synergy Platform

The central cloud is a cloud center built on the HarmonyCloud edge computing cloud platform, including highly available Kubernetes clusters, master nodes, and CloudCore. Each branch joins the cloud center as an EdgeCore, and resources are scheduled in the cloud center in a unified manner. For O&M, the cloud center enables remote branch node management, auto scaling of cloud and edge resources, and regional management of edges through different tunnels.

Prometheus federated clusters are used for monitoring. A lot of adaptations have been made for fault isolation.

Capabilities of China Mobile Online Edge-Cloud Synergy Platform

Cross-region node access: The management boundary of the central cloud is expanded to connect to edge computing resources.

Edge autonomy: Applications can run and provide services continuously even when the network connection is poor.

Resource reuse: The old servers of branches are reused to build a unified computing resource pool based on the star topology.

Technical upgrade: Branches can enjoy more efficient application deployment and O&M.

Lower costs and higher efficiency: With the advanced CI/CD pipelines of the container cloud standard delivery platform, the application O&M efficiency of branches is improved by more than 35%.

Extended computing: Computing is performed near data sources, reducing data transmission costs and central cloud storage costs.

Potential middle-end: It becomes possible to build a video service middle-end and distributed calling service system based on 5G technologies.

Achievements

Case 2: Monitoring Data Collection from Devices on a Cross-Sea Bridge

Devices that collect temperature and humidity data, or detect PM2.5 need to be connected to the board through a carrier. The board is based on the single-point system. Therefore, the board is used as the EdgeCore to connect to the entire edge computing cloud platform.

The edge computing box integrates the 5G module. The collected data is sent to the BeiDou monitoring station.

Architecture Diagram

The edge computing platform can collect data from these devices and also stream data. Some of the collected data can be directly reported. Some is processed by the MQTT broker, and then reported to the cloud database for processing. The entire edge cloud platform provides device access, data visualization, and large digital screens for edge resource management. It goes beyond a simple edge platform.

For more details about KubeEdge, visit

Github: https://github.com/kubeedge/kubeedge

Website: https://kubeedge.io/en/