Infrastructure As Code

Fowler’s compact case for treating servers like code: phoenix servers, immutable rebuilds, and why cloud without infrastructure as code leaves most of the upside on the table.

ConferenceYOW!201616:45#cloud#devtoolsPublished 2 Sept 2026

SpeakerMartin FowlerChief Scientist, Thoughtworks

Software design essayist and co-author of key agile and enterprise architecture texts.

On this page

00 · Watch
01 · Why this talk
02 · Chapters
03 · The desk cut
04 · Carry into review
05 · Transcript note

Watch the talk

Source: YouTubeWatch original ↗

Benchlog provides independent notes and commentary. The original video remains the property of its respective rights holders.

Why this talk

Sixteen minutes that still explain why most “cloud migrations” fail: teams buy elastic compute, then keep Iron Age habits — snowflake servers, login-to-fix, and undocumented handcraft.

Infrastructure as code is the real cloud unlock. Fowler’s framing — phoenix servers, config sync vs immutable rebuilds, pipelines for infrastructure — is the vocabulary every DevOps desk still uses.

Chapters

Use the timestamps to jump to the sections that matter most.

The desk cut

Watch for the phoenix-server argument and the hard rule against login-driven change. Everything else is supporting detail for those two moves.

Pull
If you aren’t using infrastructure as code, you are not going to reap the benefits of cloud.

Carry into review

Ask whether a change can be recreated from the repo alone. If the answer needs a shell history or tribal knowledge, you still have a snowflake.

Transcript note

Chapters were desk-cut from the GOTO recording transcript (YOW! 2016). Timestamps jump to the argument, not every anecdote.

The writer

Written by

Tarique Anowar

Founder & Editor · Editorial Desk

Builds Benchlog from the bench: hardware logs, software installs, and guides written after living with the stack — not the press kit.

— T. Anowar

← Previous talkSimple Made Easy Next talk →The Future of Ops Is Platform Engineering