Why we cap the number of sites per server

Cheap shared hosting works by putting a lot of sites on one server. That’s not evil, it’s how the economics work. The problem is that “a lot” stops meaning 40 and starts meaning 4,000. At that point, your site isn’t sharing a server with neighbours; it’s living in a block of flats with a broken lift and a dodgy boiler.

We cap the number of sites per server. Hard. It’s the single biggest decision behind how Cam Cloud performs, and it’s boring enough that nobody else is shouting about it, so here we are.

What a typical shared-hosting server looks like

A standard bargain-basement VPS from a provider you’ve heard of will run something like:

• 8 CPU cores
• 16 GB of RAM
• NVMe SSD

That’s a respectable machine. If you put one busy WordPress site on it, it flies. If you put 50 small sites on it, it still flies. If you put 3,000 sites on it, which major budget hosts routinely do, then the average site gets about 5 MB of RAM and a fraction of one core. That’s if everyone’s asleep. The moment two sites get a bot crawl at once, everyone slows down.

This is the point where the support script tells you to “upgrade your plan”.

The Cam Cloud maths

Our Growth tier targets ~30 sites per server. Not because we can’t technically cram more on, we obviously could, but because that number lets us promise:

• Every site has real memory headroom, not theoretical
• Background tasks (cron jobs, plugin updates, backups) don’t starve front-end requests
• One site getting hammered by a crawler doesn’t take the others with it
• We can restart PHP pools without half of Cambridge noticing

30 is the number we can sustain and still hit the 99.9% uptime target. Could a different number work? Probably. Could we go to 60 and squeeze margin? Sure, and we’d be back to making excuses.

The boring follow-through

Capping isn’t a one-off marketing promise; it’s a daily operational discipline. Every new site we onboard is checked against our current server load. If a server’s near its cap, the next site lands on a new one, even if it’s slightly less efficient for us.

It also means we can’t onboard 200 sites overnight. We don’t want to. Hosting isn’t a land grab; it’s a long game. We’d rather have a few hundred happy, quiet websites than thousands of unhappy, loud ones.

What it means for you

If you’re moving from a “£3/month” host and your site feels sluggish, it’s almost certainly not your theme. It’s the 3,000 neighbours. Move somewhere with room to breathe and the exact same WordPress install will feel like a different site.

That’s not marketing. That’s just arithmetic.