When I first heard about mainframes, one of the things that struck me funny was the claim that mainframes are “never” IPLed (rebooted, as in Initial Program Load). The development system that we use at work, however, is getting IPLed every week… ;-) Those 30-year-old mainframe newbs constantly mess something up, and a harmless submitted batch job runs amok and abends (crashes, as in Abnormal End) over and over. Just as with rebooting modern PCs, an IPL is the fastest way to get back to square zero and re-evaluate where the problem started.
But then, this mainframe is the development system. It’s a playground for DEV and QA (development and quality assurance) where abends are caught before the program’s final recut (the release build, alluding to old paper punchcards) for production.
So where are these mainframe that “never” need rebooting, doing monotonous work in the basement of some bank? How literally do I take this “never”? Surely they must be rebooted after a random power outage, or for maintainance of their parts, or… don’t know, after, say, an earthquake?
Here are some photos from an IBM data center in Tokyo containing a variety of IBM servers (mainframe and non-mainframe). Many of these IBM machines fell horizontally to the data center floor, bending frame metal and stretching cables.
However, all these IBM servers kept running, which is a testament both to the IBM engineers who designed and built them and to the IBM data center planners and managers who sweated all the little details, including leaving enough slack in the cables.
The IBM storage units kept running, too, with some concurrent error checking automatically triggered as a precaution. There were no service interruptions, and there was no need to switch over to a disaster recovery site.
Oh. So that’s what they mean. o_o