Dragonfly Reaches CNCF Graduation as P2P Distribution System Proves Critical for AI-Era Container Workloads

The Cloud Native Computing Foundation (CNCF) has graduated Dragonfly to its highest project maturity level, marking the peer-to-peer file and image distribution system as production-ready and broadly adopted across the cloud native ecosystem. The milestone caps an eight-year journey from Alibaba’s internal tooling to a foundational component of large-scale container infrastructure.

Dragonfly uses peer-to-peer acceleration to distribute container images, OCI artifacts, AI models, and other large files across clusters. In production deployments, the project claims to reduce image pull times from minutes to seconds and cut storage bandwidth consumption by up to 90 percent. At companies including Ant Group, ByteDance, Kuaishou, and Datadog, the system supports tens of millions of container launches daily.

“Dragonfly’s graduation reflects the project’s maturity, broad industry adoption and critical role in scaling cloud native infrastructure,” said Chris Aniszczyk, CTO of the CNCF, in the official announcement.

From Sandbox to Graduation

Alibaba Group open-sourced Dragonfly in November 2017. The project entered the CNCF Sandbox in October 2018 and advanced to Incubation status in April 2020. A complete architectural overhaul arrived with Dragonfly 2.0 in 2021, and the Nydus subproject, which provides on-demand container image loading, was open-sourced in January 2020.

The community growth figures are substantial. Contributor counts rose from 45 individuals at five companies to 271 individuals across more than 130 organizations, a roughly 500 percent increase. Commit activity expanded from approximately 800 to 26,000, representing 3,000 percent growth. The project now counts 1,890 total participants.

AI Workloads as a Growth Driver

The timing of the graduation aligns with a broader shift in how organizations distribute large artifacts across clusters. Generative AI workloads, which require pulling multi-gigabyte model files to potentially thousands of nodes, place particular strain on traditional registry-based distribution. Baptiste Girard-Carrabin of Datadog noted that the Nydus subproject “reduced image pulls significantly” and that “AI workloads previously taking 5 minutes now start in seconds,” according to InfoQ’s coverage.

Dragonfly integrates with Kubernetes via Helm charts and supports observability through Prometheus and OpenTelemetry. A third-party security audit was completed in 2023, followed by a joint CNCF TAG Security assessment and threat modeling exercise.

Governance and What Comes Next

Zuozheng Hu, founder of the Dragonfly project, described the graduation as “a new starting point” and signaled plans to continue expanding capabilities in data distribution. Karena Angell, chair of the CNCF Technical Oversight Committee, praised the collaboration between maintainers, TOC members, and end users in ensuring the “strength and credibility of the CNCF ecosystem.”

Dragonfly joins a roster of 29 graduated CNCF projects that includes Kubernetes, Prometheus, Envoy, and Containerd. The graduation comes as KubeCon + CloudNativeCon Europe 2026 gets underway in Amsterdam this week, where several additional project announcements are expected.