I’ve recently made https://baremetalsavings.com/ as a toy project, and I’d like to know how useful it is beyond my use case.
First, would you see this as a useful “canary in the coal mine” for reconsidering bare-metal/cloud strategy?
And second, what benchmarks would you like to see done on the cloud, to enable comparisons between different hardware? Right now I only compare instances to the closest physical hardware matching the cloud provider spec. Benchmarks will allow to expand to more options.
So I run 30ish bare metal dedicated at hetzner, ovh, and front with instances at aws, my initial feedback is, where’s the comparison the other way, how much we’d save or lose by moving to cloud from bare.
Thanks @User.Name ! The 37Signals story was indeed one of the catalysts for this.
@nathanr If you happen to own the exact some hardware that cloud providers use, then it’s quite straightforward, just add the equivalent nodes. However, this is much harder to do for any arbitrary hardware, because performance is not directly comparable.
BareMetalSavings itself runs on a Hetzner machine with Ryzen 3700X. Without bechmarking, there’s no way to tell the advantages of going to, say, AWS, because AWS doesn’t have an instance that’s comparable to a 3700X.
That’s why I want to do at least some bechmarking, to help people do exactly what you mentioned. Unlike checking for a specific use-case for a client, though, I am not very experienced with generalized benchmarking.
I thought maybe I could use Phoronix’s OpenBechmarking? Requires more research to understand if that is applicable or not. I hope people here have some advice!