Hacker News Reader

Can you explain why you need one machine never falling over ever instead of a cluster of machines never falling over at once collectively?

You don't. Either would work. Getting rid of a system running cobol on a mainframe is perhaps the hardest work possible in software engineering.