VPS vs Home Lab: Which Host Belongs Where?

The question is workload fit, not tribal loyalty

VPS vs home lab debates turn into sports teams. Operators need a calmer frame: where does each dataset want to live, what is the blast radius when power blinks, and who pays the upload bill when relatives stream photos? Many operators run hybrid home + VPS because neither location wins every workload.

Home lab strengths and honest weaknesses

Home labs win on bulk storage cost and LAN speed. Fourteen terabytes on spinning rust amortizes faster than cloud block storage if you already own the chassis. Local latency for Immich uploads from phones on Wi‑Fi is unbeatable at 300 Mbps LAN.

They lose on uplink reliability, CGNAT, residential power events, and physical theft risk. ISP upload caps (often 20–50 Mbps in the US) define remote experience more than Jellyfin settings do. Summer brownouts do not care about your ZFS pool.

Noise and heat matter in apartments. A fanless mini-PC at 15 W is fine; a transcoding tower is not.

VPS strengths and honest weaknesses

VPS hosts win on public IPv4, datacenter power, snapshot workflows, and predictable bandwidth. A $5–7/month slice with 20 GB SSD is a excellent home for WireGuard endpoint, status page, or a static blog — not for four terabytes of RAW photos.

They lose on storage pricing, GPU passthrough on a budget, and data gravity. Egress fees surprise people who treat a VPS like cheap NAS. Snapshot restore is fast; moving ten terabytes off is a project.

Split the workload — a pattern that survives

Keep photo originals at home with encrypted off-site copies; run public-facing blogs, VPN endpoints, or status pages on a VPS. Document which system of record owns each dataset in STACK.md — future-you will forget why Gitea lives in the basement but the blog lives in Falkenstein.

Example split from review-host:

• **Home (Proxmox):** Immich library, Vaultwarden, Jellyfin, local backups, bulk HDD.
• **VPS:** WireGuard relay when CGNAT blocks inbound, Uptime Kuma external perspective checks, static marketing site.
• **SaaS/hybrid:** email delivery, domain DNS, transactional SMTP.

That split keeps heavy data off expensive VPS disks while still giving you a stable front door for tunnels and monitoring.

Measure before promising remote access

Test upload speed at different times of day. A stack that works at midnight may fail during evening congestion. Design remote access around measured capacity, not brochure Mbps.

# schedule three runs: morning, evening peak, weekend
iperf3 -c your.vps.host -t 30
curl -s https://ifconfig.me # confirm which path you are testing from

If evening upload holds 25 Mbps, cap remote streams accordingly — see our media bandwidth guide. If upload is single-digit, default to VPN-only admin and LAN-first apps.

Blast radius and maintenance

Whichever path you choose, automate updates and backups first. Hosting location does not fix neglected maintenance. Snapshots on a VPS are not backups if you never restore them; RAID at home is not backup either.

We run quarterly restore drills on both sites: pull a restic snapshot from object storage to scratch, and restore a Proxmox VM from backup storage. The location differs; the ritual does not.

Cost sanity check (illustrative)

Home: hardware amortization ~$12/mo, power ~$1.50/mo for a 15 W box, drives on their own schedule. VPS: ~$5–7/mo for a tiny relay, more if you oversize RAM for Docker. Object storage for off-site encrypted backups: ~$4/mo at hobby scale. Compare against what you would pay SaaS for the same jobs — see Cost of self-hosting for a fuller worksheet.

When to start where

Start on a VPS if CGNAT blocks you, you travel often, or you need a public IPv4 this week. Start at home if data gravity is terabytes and LAN users are the primary audience. Most mature stacks become hybrid within a year — plan for that instead of forcing a purity test.

Lab note: VPS as monitoring perspective

External Uptime Kuma checks from the VPS caught a hairpin NAT failure that internal checks missed: the site looked fine on LAN, broken for everyone else. A second location is not vanity — it is a vantage point.

Related reading

See Cost of self-hosting, Secure remote access, Home lab network segmentation, Offsite backup to S3-compatible storage, and How to choose your first stack.