Project Lightning Talk: KubeEdge Everywhere: From Graduation To Global Adoption - Yue Bao

This talk introduces KubeEdge, an open source cloud native edge computing project that integrates Kubernetes with edge technologies. The speaker, Hongbin, discusses KubeEdge's architecture designed for lightweight edge computing and how it provides seamless coordination between cloud and edge environments, supporting diverse resource types such as ARM and x86 architectures. He highlights various scenarios where KubeEdge is applied, including its use in satellites operating in low earth orbit, where it processes data on the edge to mitigate bandwidth limitations. Additionally, KubeEdge supports sectors like transportation and energy by facilitating cloud-native applications and enhancing AI collaboration through its Sedona subproject. The speaker emphasizes the project's growth, with contributions from over 2,000 global developers and participation in multiple industries.

So, my name is Hongbin. So, currently I'm the technical steering committee of the KubeEdge. So, first let me give us a real How many of you heard or used KubeEdge? Please raise your hand. Uh maybe a few people. So, that's why I'm here, right? So, firstly, uh let's talk about So, what's the So, actually, KubeEdge from its name, you can see that the Kubernetes plus Edge. So,

actually, this is the first cloud native edge computing open source project. This project was introduced after 7 years uh incubating or development. It was graduated at the 2024. So, in this period, we have a more than 8,000 stars, about 3,000 forks forks on GitHub. At the same time, more than uh uh 2,000 contributors from more than 100 countries to contributed or developed this uh Uh from its

architecture, you can see the edge computing should be lightweight architecture. On the cloud side, we are still using Kubernetes. But, we customize on another features or some uh enhancement to facilities the edge scenarios. So, you can see we customize uh you know, communicating communication channel channel between cloud and edge side so that the edge side can be autonomous. At the same time, we can provide uh similar

seamless cloud edge coordination with consistent experience. That means every you know, command you are using Kubernetes can be a suitable to the edge side. And edge, we can accept low resource readiness, and we can support all kinds of architectures like the arm, x86, CPU, GPU, and other architectures. We can simplify the device communication. We provide a unified mapper to support the various protocols to support various devices.

Also, it's open system. So, many device equipment vendors can be um run a lot of some, you know, extensions by using KubeEdge. Everybody is talking about AI. So, KubeEdge has a sub group subproject named Sedona. It can enable cloud cloud and edge AI collaboration. So, actually, we can provide a unified data set and the model management across the edge and the cloud. And we can support joint

inference, incremental learning, federation learning, as well as lifelong learning. Also, we are compatible with mainstream AI framework and the provide extended interface for developers. Because KubeEdge is scenario-based, that's why it's get graduated and the more extension and attention. So, we can have more and more traditional industries to move to cloud native by using Kubernetes. Here, we list the quite a few industries who are using Kubernetes. They

are including transportation, energy, industrial, CDN, cloud native vehicles, satellite, and all other industries. So, next, I can list a typical scenarios which are using Kubernetes in their business. The first one, and I think the most famous one is KubeEdge into outer space. Low earth orbit a satellite. So, you can see, right now, there are many and the many satellites in the sky, the satellite has limited computing

computing power with limited, you know, network bandwidth. But, the more and the more data was connected by the satellite. So, how to handle this situation? So, we run satellite as edge side, edge node by using Kubernetes. So, all the data which are connected by the satellite can be handled and the processed in the sky by satellite itself. All the data you calculate the data or you process

the data can be transferred back to the ground station. The ground station has much more computing power and much more, you know, capabilities. So, this will have a satellite and ground the state ground the station, that means cloud edge collaboration. So, this can, you know, balance the satellite the resource limitation, network limitation, and the sufficient power from this from So, here we can run multi-modal collaborative between

satellite on the small models to reduce the latency, and the ground stations can use large models to facilitate the the simple identification. Also, we can have the incremental training on hard samples. We have the cloud native application management. The next scenario is a new energy vehicle powered by KubeEdge. new energy vehicles are running as edge It can handle, you know, continuous integration or various development by using

cloud native methodologies. We can support network architecture, large scalability, flexible scaling, as well as edge autonomous. So, on the right side, we list the one typical use case by using KubeEdge to predict the new energy new energy vehicle battery prediction. You can see that we can run in first port on the edge side and we can process the data transfer back to the cloud side for further

pre-modeling. So next one is multiple cloud native robots. So we can have the cloud to understand the new natural natural language to help the cloud native robots to run simultaneously. This will be have a deep dive session in another session. Okay, our goal for KubeEdge is make cloud native ubiquitous. So here we need to do the our GitHub as well as like so welcome to join. Thank

you. Thank you, sir. Little bit over time, but that is okay.