POWER your CLOUD
summary
Starts from a 2005 data center with 35 compute nodes, 15 storage nodes, 5 networks and ~10% CPU / ~20% storage utilization and describes converting that into a multi-architecture cloud. Key outcomes: migrated from VMware to KVM in two years, built an in-house PowerEL Linux distribution that runs on POWER, x86 and Raspberry Pi, and automated provisioning with Ansible + Concourse + QEMU + Podman tied to Request Tracker. Provisioning drops from ~2 hours to ~30 minutes (surveillance clusters ~45 minutes); FreeBSD/ZFS for storage and Jetson for Edge AI are used.
AI-written summary · about
chapters (12)
- 0:00Introduction to cloud journey
- 1:06Data center challenges in 2005
- 3:11Rebuilding infrastructure from scratch
- 6:18Concept of rebuildability
- 8:47DevOps philosophy and culture
- 10:55First cloud ideas and KVM migration
- 12:43High availability and clustering
- 14:29Consolidation and multi-architecture setup
- 17:47Software stack and distribution choices
- 20:13Automation with Ansible and CI/CD
- 23:03Cloud experience for customers
- 25:31Deployment speed improvements
related talks
52mLet’s build a private cloud – how hard can it be?
16mHow We Built a Cloud On K3s: The Learnings Of Growing Up Fast
57mWhat Is the Cloud?
36mOpenPOWER and GPU Porting Platform for FOSS Using OpenStack at OSUOSL
53mPrivate Cloud with OpenSource – "Now for Real!"
15mJourney to Hybrid Multi-Cloud
21mBack to the original mission: an open cloud operating system
20mDeployment Strategies for Cloud Applications: Intel Case Study