If I was (re)designing this, I would keep the existing bare metal server but I would also put in place double (or triple) cloud redundancy/failover. We all love HN so much that it should have zero downtime :-)