Observing Fastly’s Network at Scale Thanks to K8s and the Stri...

on YouTube ↗37m

" summary

Fastly describes Autopilot: a telemetry pipeline that ingests sFlow/NetFlow over DTLS from many PoPs, enriches with pmrt and Prometheus metrics, buffers in Kafka using the Strimzi Kafka operator, and exposes an in‑memory gRPC API driving routing changes. It processes hundreds of thousands of packets per second, runs on the internal Elevation Kubernetes platform with Helm, FluxCD, Harbor and Vault, solved UDP blackholing and AWS security‑rule limits with a Calico global policy plus a small garbage‑collector controller, and reached production across ~20 clusters after a four‑week nonprod rollout.

AI-written summary · about

" chapters (4)

1. 0:00Start
2. 3:50PROJECT AND PLATFORM OVERVIEW
3. 19:28DEPLOYING ON ELEVATION: EXPLORE, BUILD, DEPLOY
4. 35:01QUESTIONS?

" related talks

• 53mImplementing Kubernetes on with confidenceJeremy Proffitt, Frank Reno · AWS re:Invent 2019
• 09mKeynote: Scaling Observability Pipelines in the EnterpriseAnurag Gupta, Eduardo Silva · FluentCon North America 2021
• 39mStrangling the Legacy with KubernetesRein Remmel · Build Stuff 2019
• 35mGetting Beyond FaaS: The PLONK Stack for Kubernetes DevelopersAlex Ellis · Serverless Practitioners Summit 2019
• 14mMonitoring Network Traffic in KubernetesMichael Gerstenhaber · AWS Container Day 2019 San Diego
• 38mFlyte: Cloud Native Machine Learning & Data Processing PlatformKetan Umare, Haytham Abuelfutuh · KubeCon + CloudNativeCon North America 2019
• 35mAutoPiloting Kubernetes: how we built autonomous Control PlanesDario Tranchitella, Gianluca Mardente · ContainerDays Hamburg 2026
• 38mA Deep Dive into FinOps Integration with Kubernetes for EnhancedFrancesco Lumpp · Incontro DevOps Italia 2024