Hetzner now throttling shared CPUs if you use >20%
In this technical retrospective (2024), @micw shares practical architectural lessons learned from running, scaling, or migrating production software under lean constraints.
HN engineers debated the tradeoffs between cloud convenience and raw server economics, database indexing strategies, and resisting premature architectural complexity.
Where does it say that? Here is exactly the page I see: https://web.archive.org/web/20240912104240/https://docs.hetzner.com/cloud/servers/faq/
Plans (CCX) with dedicated vCPUs have their CPU resources exclusively. Here one vCPU equals one thread of a physical CPU core. Instances with dedicated vCPUs offer continuous and predictable high CPU performance. We recommend them for systems with high production loads and CPU intensive applications. They run on high-performance hardware and come with a generous allocation of I/O and network power.
Of course that implies there is some throttling. But I see nothing about a hard 20% cap.
I would think if you have something that needs >20% CPU constantly, you really shouldn't be relying on a shared resource for it in the first place.
To put it in relation we are talking here about less than 15€ per month for 2 dedicated vCPUs/8GB/80GB vs. 5€ per month with half the RAM and HDD. It really doesn't matter.
What if you needed to use your hetzner or Amazon computers to do………. computing?
Hetzner has changed their FAQ for shared CPU cloud servers. Similar to AWS they define a CPU baseline (20-33%, depending on server type). If that baseline is permanently exceeded, the CPUs are throttled.