37signals is completing its on-prem move

theregister.com

57 points·xrayarx··8 comments

Discussion earlier: https://news.ycombinator.com/item?id=43936754

They say that the move didn't require adding any staff, but aren't all of these physical servers going to require somebody to do occasional physical tending to them? What happens when a server fails? Sure the software might be able to handle failover and absorb the load, but isn't somebody going to have to go deal with the failed server at some point?

Dhh has addressed this a dozen times - they shipped the servers to some datacenter which manages everything for them. And they, obviously, have hardware redundancies

Colocation, essentially.

Colocation services are decades old and exist all over the world. Did you never hear of them?

hello,

imho. as always ...

its not only colocation of (server|network|...)hardware.

its possible to outsource your infrastructure to a business which "manages" it for you.

idk. think ibm, but also many other larger & smaller vendors who offer such services :)

just my 0.02€

Even if they don't use colocation, is there a business difference between hiring dozens of AWS engineers vs contracting dozens of infra engineers?

This can be mostly handled in the planning. Plan for server failures, but datacenters will also offer 'remote hands'. You can store a bunch of spare parts in one of your racks (see rack-mount-drawers), then call up the DC, and ask them to swap component A in server B with spare C in drawer D.

Reply on news.ycombinator.com