"From On-Prem to the Cloud and Back Again"
summary
Claims: FastMail runs about 250 servers across three datacenters storing ~5 PB for ~200k paying customers, operates ~100 services with 5–10 prod deployments a day and 1–5 incidents weekly, staffed by four platform engineers plus five contributors. Practical tooling: Proxmox, VyOS, WireGuard, PXE/FAI imaging, Unbound/Knot DNS, Podman over Docker, and a caution about distributed KV and Kubernetes. Concrete advice: automate server bring-up and deployments, standardize on boring well‑known tech, and avoid undercooked distributed state.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:08Who this talk is for and platform definitions
- 5:24Building a platform
- 8:12Cloud vs on-prem trade-offs
- 13:25Disadvantages of self-hosting and cloud advantages
- 16:17Starting with self-hosting
- 18:50Learning with limited resources
- 20:57Platform building best practices
- 23:46Tools to avoid and recommended
- 26:16Change management practices
- 27:21Software recommendations
- 28:25Closing and Q&A
related talks
28m10 Billion Messages in a Day, How We Built the Infrastructure That Delivers
21mHybrid Cloud - nothing to worry about! 🌥️
26mCloud vs. Control: Save a Business by Deploying Global Applications w/ Local Hardware
44mEmail at Scale: How We Survived 800M Emails/Year (and DNS)
37mTo Cloud or Not to Cloud?
43mFast and Furious: Startup scaling with Kubernetes
34mLessons Learned from Five Years of Multi-Cloud
56mGetting all the 99.99(9) you always wanted