The Self-Hosting Responsibility Spectrum

February 13, 2026

Not all self-hosting is equal. The difference isn’t hardware, its operational accountability.

I took good look at my responsibility matrix and decided it was time for an upgrade to clarify where systems sit and what it all really means in practice:

Level	Environment	Hardware Owner	OS & Services	Backups & Recovery	Purpose
0	Saas	Vendor	You	Vendor	Consumption
1	Hosted Platform	Vendor	You (apps)	Vendor	Partial control
2	VPS/Cloud VM	Provider	You	You	Infrastructure responsibility
3a	Full Stack Homelab	You	You	You	Skill development
3b	Full Stack Production/Critical	You	You	You	Reliability first

Now let’s talk about what each level implies.

Level 0 – SaaS: Consumption

You configure settings while the vendor handles infrastructure, uptime, security patching, backups and disaster recovery. If you can’t explain where your data is located, backed up, and how to export it – you’re trusting marketing, not architecture.

Common mistake: Assuming SaaS means “no responsibility.” You still own identity, MFA enforcement, and access control – and maybe even things like compliance if we’re talking about business.

Level 1 – Managed Platform

You control the application layer; the provider manages the OS, runtime, and infrastructure. Ask yourself – if the provider had a breach tomorrow, what would you be accountable for?

Common mistake: Believing “managed” means “secure.” You still have/own misconfiguration risk.

Level 2 – VPS/Cloud VM: True self-hosting begins here

The provider guarantees compute availability. Everything else is on you. If you don’t have monitoring and tested backups, you’re renting risk and not hosting infrastructure.

You manage:

• OS patching
• Firewall rules
• Service configuration
• Backups
• Monitoring
• Incident response

This is self-hosting - even if the server lives in a datacenter across the country. For example, the VMs I operate in New York via a cloud provider are no less self-hosted than the hardware in my house. No one else is patching them. No one else is backing them up. If they fail, I fix them.

Common mistake: Treating a VPS like a disposable platform – the moment it runs your day, it becomes production.

Level 3a – Full Stack Homelab: Intentional learning environment

A real lab should be rebuildable from documentation. If you can’t wipe it without fear, it’s not a lab - it’s production in denial.

You own everything:

• Hardware
• Networking
• Hypervisor
• Operating systems
• Services
• Backups
• Recovery

This is where experimentation lives – intent is the defining characteristic.

In my case, that includes:

• A 3-node Proxmox mini-cluster
• A Proxmox Backup Server
• Raspberry Pi lab nodes for clustering and Kubernetes testing
• “Zoltan!” - a wipe-at-any-time test host

These systems exist so I can break things safely and learn from them.

Common mistake: Letting lab systems quietly become critical dependencies without upgrading operational discipline.

Level 3b – Full Stack Production: Reliability first

This is where many people blur the line. You still own everything - but the purpose shifts from experimentation to stability.

In my network, this includes:

• My inference system
• Synology storage
• 45 Drives Pro 4 backup target
• Any infrastructure that directly impacts income or daily workflow

These are self-hosted - but they are not Homelab systems. The word “lab” implies safe failure. Production systems don’t get that luxury. If downtime affects your income, reputation, or irreplaceable data, you are running production infrastructure - whether you admit it or not.

Common mistake: Continuing to treat production systems casually because they live at home.

The Media Server Question

Self-hosting a media server is great. It teaches basic networking, storage management, and service configuration.

But ask yourself: what’s the primary goal?

• Entertainment? or;
• operational growth?

If the system exists mainly to stream movies, it’s a home server. That’s fine. But unless you’re intentionally designing for resilience, monitoring, and recoverability, it’s not a Homelab.

A lab implies experimentation and measurable skill development. Convenience alone doesn’t qualify.

The Risk and Learning Overlay

Environment	Operational Risk	Learning ROI
SaaS	Low	Low
Managed Platform	Medium	Medium
VPS / Cloud VM	Medium–High	High
Full Stack Lab	Controlled High	Very High
Full Stack Production	High	Moderate (stability-focused)

• A lab should maximize learning while containing risk.
• Production should minimize risk while applying what you’ve learned.

Confuse those two, and things get expensive.

Why This Matrix Really Matters

When people say, “I run a Homelab,” what they often mean is “I own hardware.” That’s not the same thing.

The meaningful distinction isn’t physical location. It’s:

• Operational responsibility
• Intent
• Risk tolerance
• Purpose

From a Slackware box in the late 90s to modern Proxmox clusters and cloud VMs, the lesson has stayed consistent:

Owning hardware is easy. Owning responsibility is the point.