[Beginner] · The roles that own the physical and platform layers, what each actually requires, and the realistic routes in — including from where you already are.
The previous four notes describe the machine, the building and the wiring. This one is about working on them, because “data centre” covers several genuinely different jobs.
The roles, and what separates them
| Role | Owns | Core skill | Ceiling |
|---|---|---|---|
| Data centre technician | Racking, cabling, hands-on remediation | Physical, methodical, follows procedure | Modest — but a real door in |
| Network engineer | Switching, routing, the fabric | Deep networking | High; specialised |
| Systems / infrastructure engineer | Servers, OS, virtualisation, storage | Linux, automation | High |
| Cloud engineer | Cloud infrastructure as code | Terraform, one cloud deeply | High, and the largest market |
| SRE / platform | Reliability, delivery platforms | Software + operations | Highest → 12 |
| Hardware / capacity | Procurement, power, cooling, planning | Electrical + facilities | Specialised, scarce |
The split that matters: hands-on-metal versus infrastructure-as-software.
Hands-on roles — technician, cabling, hardware — are geographically bound (you must be at the building), pay less, and have a lower ceiling. They’re a genuine entry point with few prerequisites, and a poor destination.
Infrastructure-as-software roles — cloud, SRE, platform, network automation — are remote-friendly, pay well, and are where the field has moved. This is the direction to aim at, and it is reachable directly.
The honest market picture
Physical data centre work is growing — AI capacity build-out is enormous, and there’s real demand for technicians, electricians and facilities engineers. But it is local work: you must be near a data centre, and the big ones cluster in specific regions.
Cloud and platform work is the larger, remote-friendly market, and it doesn’t require the physical layer as a prerequisite — only as understanding, which is what notes 01–04 are for.
For you specifically: the remote-work goal in your catalogue points clearly at the second column. Understand the physical layer; work on the software layer. Notes 01–04 exist so that “availability zone” and “IOPS” and “oversubscription” mean something physical when you’re reasoning about a cloud bill or a latency graph — not so you can rack a server.
What you actually need
Non-negotiable, in this order:
1. Linux, deeply. Not “I can cd” — processes, systemd, networking, storage, permissions, logs, troubleshooting a box that won’t boot. This is the single highest-return skill in infrastructure and there is no way around it → 01-linux, and the RHCSA track is a genuinely good structure for it.
2. Networking. Subnetting, routing, DNS, TLS, firewalls, and the ability to debug why two machines can’t talk. The most common gap in self-taught infrastructure people → networking.
3. One cloud, properly. Not three shallowly. AWS has the largest market; Azure dominates enterprise; GCP is strong in data/ML → 08-cloud.
4. Infrastructure as code. Terraform, and configuration management. The line between “sysadmin” and “infrastructure engineer” is essentially whether your infrastructure is in git → 07-IaC.
5. A scripting language. Python and Bash → Python · Bash.
6. Containers and orchestration. Docker, then Kubernetes → 03-docker · 10-orchestration.
Then: observability → 07, CI/CD → 05, and security posture → cloud security.
Certifications — where they actually help
More than in software development, and this is a real difference between the fields. Infrastructure hiring uses certs as a filter far more than dev hiring does, particularly for a first role and particularly outside a strong personal network.
Worth it:
- AWS Solutions Architect Associate — the broadest signal, widely recognised
- RHCSA — proves real Linux ability, hands-on exam, hard to fake. This vault already has a track for it → RHCSA
- CCNA — still the networking credential, if you’re going that direction
- CKA — Kubernetes, hands-on, respected
The honest caveat: a cert gets you past a filter. A cert with nothing behind it fails the interview, and infrastructure interviews are unusually practical — you will be asked to debug something. Pair every cert with something you actually built.
The realistic routes in
From software development — your position, and the shortest path. You already write code, use git, and understand systems. Adding Linux depth, one cloud and Terraform makes you a cloud/platform engineer, and the software background is an advantage, not a starting deficit: the field has moved toward infrastructure-as-software, and most people in it came from operations and had to learn to code. You already did the harder half.
From help desk / IT support. The classic route: support → sysadmin → infrastructure → cloud. Slower, and it works.
From a technician role. Genuinely available with few prerequisites, and the way to use it is to automate your own job and move sideways into engineering.
From a degree. Systems engineering is directly relevant — capacity, power, cooling, reliability and interfaces are the discipline → systems engineering.
What to build
Infrastructure is proven by what you’ve run, not what you’ve read, and the vault’s standing position applies here more than anywhere → reading is not a rank.
- A real VPS, doing something real. Buy one, harden it, deploy something, keep it up → 04-vps. Publicly reachable and actually used beats any lab
- Everything in Terraform, in a public repo. VPC, subnets, security groups, an instance. Destroy and recreate it from scratch — that’s the claim IaC makes, and doing it proves you understand it
- A CI/CD pipeline that deploys on push → 05-ci-cd
- Monitoring, with an alert that has actually fired → 07
- A small Kubernetes cluster — k3s on cheap VPSes teaches more than a managed control plane, because you have to fix it
- A home lab, if hardware appeals — an old machine, a hypervisor, some VMs → virtualisation. This is where the physical notes stop being abstract
Write up what broke. An infrastructure portfolio is post-incident notes more than it is repositories — the ability to describe a failure clearly is most of the interview.
Related
- devops — the whole track, in order
- the physical layer — notes 01–04
- SRE and platform engineering — where the ceiling is
- networking — the most-skipped prerequisite
- the Primetechie path — the gated progression
Source: [reference] — written Aug 2026; cross-referenced against the roadmap.sh devops and network-engineer roadmaps.
Software engineering
Infrastructure & security
Computer science
Hardware & engineering