2022 Superuser Awards Nominee: Ant Group

The Kata Containers team is part of the infrastructure department of Ant Group. The team is led by Xu Wang, who co-announced the project in Austin in Dec 2017, and the team consists of more than 30 members. The Kata team is responsible for the container-related R&D of Ant Group and working for the sustainable IT infrastructure with Kata Containers in Ant Group. Many of the team members have been active upstream Kata Containers project contributors and maintainers since the first day of the project, such as Peng Tao and Fupan Li.

Ant Group is a financial technology company aiming to help everyone enjoy inclusive, sustainable financial services. In April 2022, Ant Group announced that it achieved carbon neutrality in 2021 by reducing 29,591.48 tons of CO2 emission with green computing technologies. One of the key technologies is to increase the efficiency of infrastructure by running heterogeneous workloads together and isolating them with Kata Containers.

Thanks to the strong isolation of Kata Containers, we can deploy online applications and batch jobs together on thousands of nodes without significant interference. By adopting Kata Containers and other related technologies, we have reduced half of the per-payment energy consumption in the recent Double-Eleven e-Shopping Festival, compared to three years ago.

We are one of the two founding teams of Kata Containers. Since its first launch in December 2017, we have been active contributors to the Kata project. We contributed to the community in many ways such as organizing events, drafting roadmaps, leading feature developments, enriching documents, fixing bugs as well as helping other users to adopt Kata Containers in production.

Besides Kata Containers, we also use a variety of other open source technologies in the infrastructure, such as Kubernetes, Prometheus, etcd, containerd, runc, dragonfly, Nydus, Linux, KVM, etc. Together, they helped us to build up a stable and sustainable IT infrastructure.

We run our Kubernetes clusters on tens of thousands of nodes. The size of a single Kubernetes cluster varies from a few hundred to as large as more than 10 thousand nodes.

We operated on a large scale and encountered many kinds of challenges. We have overcome and solved the scalability of many infrastructure components, upgrades, performance, and performance jitters, etc.

We are pioneer users of Kata Containers. We deploy Kata Containers on a large scale in production and innovate with real-life use cases. In the past, we introduced the shimv2 architecture and the rust version of the kata-agent. We put these features in production at a very early stage. It helped validate and mature them quickly.

Right now, we are adding a rust version of the Kata runtime with an integrated built-in VMM, to further reduce Kata’s memory overhead, and to bring in true cloud-native virtualization. It aims to be a major feature of the incoming Kata Containers 3.0 release.