Wendell Beckwith
Redfin, Technical Lead, Cloud Orchestration
Actions
As a Tech Lead for Redfin’s Compute Orchestration team, Wendell Beckwith manages all things Kubernetes and steers the adoption of cloud-native solutions. A technical reviewer for GitOps and Kubernetes, he recently presented on navigating the journey from Terraform to Crossplane at Kubernetes & Cloud Native North Dallas. Currently, he is focused on optimizing enterprise cluster efficiency and scaling Kubernetes in multi-cluster environments with Cilium, Crossplane, and Dragonfly.
Every Node Pulls Alone: Fixing Container Image Distribution with CNCF Dragonfly
Every time a Kubernetes node starts, it downloads every container image it needs from the registry, alone. With autoscaling and spot capacity creating thousands of fresh nodes a day, that means slow cold starts, a growing registry bill, and a single point of failure for every scale-up.
CNCF Dragonfly, a graduated project since January 2026, fixes this with peer-to-peer distribution: nodes share image layers with each other, and the registry serves each layer about once. We rolled it out across 30+ EKS clusters. On day one it looked great. Months later we found that only 1.4% of pulls on one of our biggest clusters actually went through it, while every dashboard stayed green.
This talk explains how image pulls work, how Dragonfly changes them, and the eleven things that broke in our rollout, from a boot race to a cache that pulled its own dependencies through itself. You'll leave able to measure how much of your image traffic really goes through your cache, and to gate new nodes until it's ready.
When AI Agents Break the Registry: How Redfin Used Dragonfly to Handle a 140TB Image Explosion
The AI boom isn't just a compute problem—it’s a data distribution crisis. As developers at Redfin rapidly adopted Agentic processes and tools, our container registry traffic ballooned from 50 TB in early Q3 2025 to a staggering 140 TB by mid-Q1 2026, causing costs to skyrocket. We knew we needed an image distribution solution, but moving CNCF Dragonfly from a basic tutorial to a real-life production environment was unchartered territory.
In this lightning talk, we'll share the high-level strategy Redfin's Compute Orchestration team used to tackle this explosion. We'll show a snapshot of our custom Datadog dashboard revealing how Dragonfly achieved an incredible 86.4% cache hit rate—shifting the data burden to peer nodes instead of fetching from remote registry providers. Attendees will walk away with hard data proving how P2P distribution directly slashes egress, cross-zone transfer fees, and deployment times when AI hits production.
Dragonfly IRL: The (missing) Enterprise Production Guide
The Dragonfly tutorials get your feet wet. They do not get you swimming in production.
When Rocket acquired Redfin and launched the Great American Home Search — a $1M giveaway featured in their 2026 Super Bowl ad — our platform saw what happens when a traffic surge meets AI-driven development workflows at the same time. Registry image pulls exploded from 50TB to 140TB. Costs followed.
We deployed Dragonfly on EKS — multiple Helm releases, per-AZ seed peers, Karpenter node pools, Gatekeeper carve-outs. We hit silent failures and we built validation tooling to prove P2P was actually running. Result: 86% cache hit rate in production.
This talk is the guide we wish we had. Full architecture, the silent failures you'll hit on day one, and the observability layer we run today. Attendees leave with a public repo: Terraform modules, Helm values, validation scripts, and Datadog dashboards.
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top