How does that solve the downtime issue? In my experience company-run instances tend to go down just as often, if not more often.