Due to the diversity of requires, solutions of Edge computing have a high degree of freedom and customization space.The trend of Hyper-Converged Infrastructure(HCI) in next generation data centers is expanding from Cloud to Edge. Traditional HCI systems provide software-defined unified solutions for computing, networking, and storage, and rely on virtualization. In the future, HCI will also integrate AI and Blockchain. A more efficient, flexible, and lightweight approach is preferred as the still limited resources at Edge when compared to Cloud. This proposal focus on rethink eBPF-based redesign and re-implementation of HCI at Edge for ARM + Linux, and comes with the following sub-topics:
1) eBPF-based In-Kernel services;
2) Ray and Redis based distributed scheduling, data processing and AI;
3) Kata containers, lightweight Kubernetes solutions like K3S & K2S, project Cilium, etc;
4) eBPF-centric new HCI solution for Edge computing.
This talk is for people who are interested in innovative technologies that have already or will greatly change the Linux-centred Edge Computing, or who are paying attention to the Hyper-Converged Infrastructure.
Some preliminary implementations or experiments have been published or gradually being made available at github. it is promised that all the code involved in our approach will be fully open sourced.
