CERN's Move from RHEL 10 to Debian 13: When Hardware Baselines Force Migration

The world's largest particle accelerator control room must replace the operating system on over 2,200 devices by year-end. The trigger is a hardware baseline: the new RHEL generation requires CPUs supporting the x86-64-v3 instruction set, and two-thirds of the machines in the facility fall short. CERN's answer is Debian. Behind this move lies a question every Linux user will eventually face: how to balance chasing the new with preserving the old?

1. Pushed Out, Not Over Money

The migration covers CERN's accelerator control systems — the infrastructure that steers particle beams — encompassing more than 2,200 machines. According to an InfoQ deep-dive published September 7, the target is to complete the switch to Debian 13 by the end of 2026.

Why leave? The unavoidable reason: Red Hat's next-generation RHEL (the 10 series) raises the hardware baseline to x86-64-v3. This instruction-set threshold demands CPU support for AVX and SSE4.2, instructions that only became widespread around 2010. While that sounds modest, for an accelerator facility running hardware for over a decade, two-thirds of the devices fall below this line.

The machines simply cannot install the new OS because they are too old. Cost is secondary.

These control devices prioritize stability. As long as an accelerator controller runs reliably, no one wants to touch it. By raising the baseline, Red Hat effectively told CERN: either replace two-thirds of your still-functional machines, or don't use RHEL 10.

CERN chose a third path.

2. Why Debian

Could they switch to another commercial distribution? CERN picked Debian for concrete reasons.

Debian 13's hardware baseline remains at x86-64-v2. The older machines that Red Hat effectively sentenced to obsolescence can run Debian without issue. That single fact explains the choice.

But there's more. Debian is a pure community project with no commercial entity that can unilaterally change direction, raise baselines, or modify source code. For a research facility that must operate for decades, "no one can single-handedly change the rules of the game" is itself a value. Today Red Hat raises the hardware baseline; tomorrow who knows what else it might raise.

Another practical factor: CERN already has extensive internal Debian experience. Many scientific software packages and detector control systems already run on Debian. Migrating there means returning to familiar territory.

3. Why Red Hat Couldn't Keep Them

CERN isn't the first to leave. The line of departures stretches back five years.

Late 2020: Red Hat announced early end-of-life for CentOS Linux. CentOS 8 would die at end of 2021; CentOS 7 would last until mid-2024. Tens of thousands of server users lost their free "production-grade" system overnight, left with two options: pay for RHEL or move to CentOS Stream as de facto beta testers.

Aftermath: CentOS founder Gregory Kurtzer created Rocky Linux; CloudLinux created AlmaLinux — both aiming to catch the abandoned users.

June 2023: Red Hat tightened public access to RHEL source code, restricting it to paying customers. This move hurt Rocky and AlmaLinux, blocking the public "rebuild" path.

From CentOS end-of-life, to source-code restrictions, to the x86-64-v3 baseline, every Red Hat step in recent years has pushed users who want "free, stable, compatible with old hardware" toward the exit.

CERN is simply the latest — and most prominent — name on that list.

4. Chasing New vs. Preserving Old

This forces us to rethink a core question: how should Linux "evolve"?

Red Hat raising the hardware baseline follows a logic of progress: new instructions bring performance gains; dropping historical baggage reduces maintenance burden. That logic is sound.

Debian holding at x86-64-v2 follows a logic of compatibility: letting a machine from ten or more years ago still install a modern system with security updates. That logic is also sound.

The article compares the two approaches across four dimensions:

Hardware baseline: Red Hat lineage steadily raises it (RHEL 10 requires x86-64-v3); Debian holds at x86-64-v2.

Old devices: Red Hat — cannot install new versions, forced hardware replacement; Debian — old machines install and keep running.

Roadmap logic: Red Hat — pursue new, shed legacy; Debian — compatibility, preserve installed base.

Governance: Red Hat — commercial company led; Debian — pure community.

There is no right or wrong, only trade-offs. The real world often needs both answers simultaneously.

For typical Ubuntu users, the answer was already written. Ubuntu takes a middle road: LTS for stability, HWE (Hardware Enablement) for new hardware. Want stability? Pin to LTS and an old machine lasts years. Want new? Switch to HWE kernel and follow hardware cadence every six months.

Whether old devices should be abandoned is not a technical question — it's a values question.

Red Hat graduates old devices; Debian extends their life. Which is right depends on where you stand. CERN voted with 2,200 machines.

That aging laptop you code on, the five-year-old NAS, the server in the rack that hasn't reached end-of-life — when a distribution raises its hardware baseline, what will you do?

What's the oldest Linux device you still run, what OS does it use, and can it still hold on?