By Andy Shi
We are very happy to announce that OpenYurt is accepted into CNCF sandbox.
In the era of IoT, more and more users prefer to process data at the edge instead of transferring it back to the data center. By doing so, users can save both time and networking costs. However edge computing brings its own challenges.
The major difficulty is with networking, and it comes in two-fold. The first challenge is quite obvious: The networking from edge to data center is less stable than that inside the data center. The current Kubernetes health check and synchronization mechanism assumes stable networking. The second challenge is the operational complexities. For example, most data centers have ingress firewalls.
So how can we ensure the connections initiated by edge worker nodes to the API server will not get rejected? A more complicated situation is how can we support edge workload lifecycle management, rolling updates and auto scaling when the workloads are scattered over different geo locations? These challenges are beyond the scopes of upstream Kubernetes.
From ACK@edge to OpenYurt
At Alibaba cloud, we seize this opportunity and offer the market ACK@edge, an edge computing product built on top of Kubernetes. It enhances the edge and data center synchronization. Its main features include: Work node autonomy to deal with the unstable networking. Edge tunnel to ensure secure and reverse communication between the edge worker node and the data center. ACK@edge has been widely adopted into different use cases such as video streaming, AI and IoT data processing.
To benefit the community, a little while ago we decided to open source the core technology of the ACK@edge product to the public. And that’s how OpenYurt was born. We are committed that the commercial product will use the same upstream OpenYurt code base.
Recently, CNCF has adopted several edge computing projects into its inventory. That’s an indication that cloud native edge computing is gaining momentum. Compared to other projects, OpenYurt is 100% compatible with upstream Kubernetes. It doesn’t change any upstream code. Instead, all the features are added on to the upstream Kubernetes. So users can easily convert back and forth between an OpenYurt cluster and a regular Kuberntes cluster. Another difference is an OpenYurt cluster has control planes in the data center and (some) worker nodes at the edge. The edge worker nodes can be either X86 or ARM architecture. But they are normal computer servers nonetheless.
In the near future, we will be building OpenYurt into a full platform, including a GUI. Another milestone is to better support AI workloads by integrating with popular AI workflows such as kubeflow.
If you are interested in these features or edge computing in general, please join us! We can’t be happier to have you. Visit the official website of our project to learn more: http://openyurt.io/en-us/. Together let’s take OpenYurt to the next level!
The views expressed herein are for reference only and don’t necessarily represent the official views of Alibaba Cloud.