> Also, Hardware isn't as hard as people make out, you install it once and it runs; if it fails there's a special out-of-band manager built in so you can remote

> Also, Hardware isn't as hard as people make out, you install it once and it runs; if it fails there's a special out-of-band manager built in so you can remotely debug it, datacenters have remote-hands personel. If you're "cloud native" you've already architected your stack to be horizontally-scalable and distributed (partly because the cloud is so expensive that running at peak load capability all the time would bankrupt you) so hardware failure should matter even less.