They achieve all of this using 100% open-source infrastructure. If I remember correctly, it’s all running on Codeberg-owned hardware as well, not some rented servers.
A company at which I once worked built a functioning server into the frame of a motorcycle. It was after I left, so I’m not sure of the details, including whether it had to be plugged in; but regardless, they called it “the world’s fastest server!” and I think that’s pretty funny.
To be fair MS makes orders of magnitude more money and has the benefit of operations at scale. Whereas codeberg’s operational budget for 2025 was 100k euro and they still need to deal with DDoS and bot scraping. They also were running off a single server up until sept’25 when they had two donated hardware services which are now hooked up to make a 3 node ceph cluster.
Meanwhile, over at Codeberg: https://status.codeberg.org/
They achieve all of this using 100% open-source infrastructure. If I remember correctly, it’s all running on Codeberg-owned hardware as well, not some rented servers.
https://codeberg.org/about
They were down for like entire day once because they moved that server to a new location by train. In a backpack.
I am disappointed. A few servers have been moved via train and stayed online. Codeberg should do better.
If that was their only downtime that year, that would have resulted in 99.7% uptime.
A company at which I once worked built a functioning server into the frame of a motorcycle. It was after I left, so I’m not sure of the details, including whether it had to be plugged in; but regardless, they called it “the world’s fastest server!” and I think that’s pretty funny.
Migrations should always incur downtime
“Hey we’re migrating, take a break for a week”
To be fair the number of users they serve is probably orders of magnitudes lower.
To be fair MS makes orders of magnitude more money and has the benefit of operations at scale. Whereas codeberg’s operational budget for 2025 was 100k euro and they still need to deal with DDoS and bot scraping. They also were running off a single server up until sept’25 when they had two donated hardware services which are now hooked up to make a 3 node ceph cluster.