So, the "unofficial slogan" — you make me laugh! But bon, it is true, no?
My OVHcloud, the first time I am put twenty sites, the RAM it is cry. Very fast. Now I am use the old Contabo box for the heavy ones, the new for the small.
So, the "unofficial slogan" — you make me laugh! But bon, it is true, no?
My OVHcloud, the first time I am put twenty sites, the RAM it is cry. Very fast. Now I am use the old Contabo box for the heavy ones, the new for the small.
This is why I keep a $5/month Contabo box as a wireguard endpoint. Traffic comes in clean, tunnels home, ISP sees nothing but UDP 51820. My playbook for this is public if anyone wants it.
SNI is table stakes. The real fun starts when you need client cert auth per vhost. nginx handles it, Caddy handles it, Apache handles it badly.
Run the numbers on a NUC drawing 15W vs a Contabo VPS for three years. The breakeven is closer than you think, and that's before you value the RAM and disk you actually control.
IPv4 scarcity is the real tax here. Single IP at home, single IP on a budget VPS—same problem, different landlord.
So now I need two boxes to do what one did in 2003, plus a VPN, plus playbooks. Progress.
Non, the electricity in France it is not cheap. I am calculate, the VPS it is win. But the fiber, the upload, this it is why I am stay at home.
The OVHcloud, the CPU steal, I am say already. The Contabo, it is better for the heavy.
You needed two boxes then as well. One died, you had no redundancy, no snapshot, no "terraform apply" to rebuild in four minutes. You had a weekend of driving to the datacenter or begging a friend with a key.
The NUC plus VPS model is the same cost spread differently, with actual recovery options.
Kiddo I don't need your ansible fanfiction. I've got a bash script from 2017 that does the same thing in twelve lines and doesn't pull fifty galaxy roles to get there.
The NUC brigade always forgets the ISP SLA. Residential fiber is "best effort" with a wink. My Contabo node has had three planned outages in five years, all announced two weeks ahead.
I did. My kill-a-watt says 23W at the wall for the i5 NUC, plus the modem, plus the switch. Call it 35W continuous. At Portland General Electric rates that's roughly $3.50/month. Contabo VPS with comparable RAM is $4.99. The gap is noise.
But the NUC can't live in a datacenter with redundant power and someone else's fire extinguisher. That's the actual value prop.
And when your ISP changes your IP, that script does what exactly? My playbook handles the API call, the DNS validation, the cert renewal, the nginx reload. Your twelve lines grow or they rot.
This. Following.
Try running your inference workloads on that budget. My 3090 pulls 350W and I still tunnel the web frontend through a $3.50 Contabo slice. The GPU does LLM, the VPS does SSL termination and SNI routing. Best of both worlds, nobody hits my home IP directly.
Caddy on the edge, Traefik on the inside. Automatic HTTPS at both layers because life's too short for certbot cron jobs.