Dragonfly V2.5.0 - Intro, Updates, Data Distribution in AI Infrastructure
Talkat KubeCon + CloudNativeCon + OpenInfra Summit + PyTorch Conference China 2026 · 2026-09-07..09-09·
summary
Dragonfly is a P2P-based system for file distribution and image acceleration in cloud-native environments; the talk covers the 2.5.0 release with enhanced security and new features for model distribution. It shows using Dragonfly to preheat and distribute AI models packaged as OCI artifacts into read-only Kubernetes volumes for faster deployments, and introduces P2P-based state snapshot and restore for AI agent scenarios.
AI-written summary · about
related talks
27mDragonfly: Make Image Distribution Efficiently and Safely in Cloud Native
43mIntro: Dragonfly
29mPull Once, Scale Everywhere: Fixing Image Pull Bottlenecks With Dragonfly
31mBuilding the Dragonfly Operator from Scratch
34mThree Approaches to Speed up Image Distribution in Cloud Native Era
34mThree Approaches to Speed up Image Distribution in Cloud Native Era
38mFlyte: Cloud Native Machine Learning & Data Processing Platform
17mDevOps for Private AI: A Scalable Platform Using Kubeflow