i have one fat always on box in a cloud provider thats costing me a fortune because its basically never idle, so im paying cloud premium for zero elasticity benefit. thinking of buying hardware and putting it in a colo. i know the sticker math favors colo hard for steady workloads but i also know the hidden costs, remote hands, a dead drive at 3am, no instant replace. for people who actually did this, did the savings survive contact with reality?
for a steady always on box the colo math is real and it survives, but only if you buy two. a single colo box means you personally are the failover at 3am. with a spare and a tested restore the savings are huge versus cloud premium on a workload that never scales down.
did exactly this for a box like yours. savings were real but the first six months had a learning tax, a bad stick of ram, a flaky nic, stuff the cloud hid from me. once it was stable it just ran. budget for that rough patch and it is worth it.
whatever you do, your backups and restore have to be bulletproof before you move, not after. in cloud you get lazy because snapshots are one click. in colo a dead drive with an untested restore is your whole business gone. test the restore, not just the backup.
two boxes and a tested restore is the recurring theme, which makes sense. i was pricing one box against one cloud instance and that is the wrong comparison. redoing the math with a spare included and it still wins, just by less. barry the restore point is well taken.