> Maybe there's some forms of clever caching involved? That could explain why the monthly expenditures seem to be moving around so much, which is unusual. Do you have a link about it?
Mostly it is just better/newer hardware for the database servers - both CPU & NVME storage. But also more hosts, that are not sharing system resources, and have sub-ms latency between each other.
Yes, costs change as hardware is added. But this leads to another advantage which is hard to understate:
It is very easy for Codefloe to adapt its hosting architecture because everything is managed with infrastructure as code. So as bottlenecks are found, improvements are benchmarked, change can be deployed safely and reliably. Codeberg is much riskier to change at this point.
Some of this is public here[1] - but I don't really expect it to convince you. What should convince you, is simply using Codeberg. It feels like being on dial-up.
Mostly it is just better/newer hardware for the database servers - both CPU & NVME storage. But also more hosts, that are not sharing system resources, and have sub-ms latency between each other.
Yes, costs change as hardware is added. But this leads to another advantage which is hard to understate:
It is very easy for Codefloe to adapt its hosting architecture because everything is managed with infrastructure as code. So as bottlenecks are found, improvements are benchmarked, change can be deployed safely and reliably. Codeberg is much riskier to change at this point.
Some of this is public here[1] - but I don't really expect it to convince you. What should convince you, is simply using Codeberg. It feels like being on dial-up.
[1] https://pat-s.me/codefloe-launch/#takeaways-from-my-time-wit...