The veteran capacity manager has stared down the maw of disaster, laughed defiantly, and his career lives on. The hard limit of his core data operation is how much can be run by the diesel electric system when the grid goes down.

Juxtapose the power limit against hardly imaginable performance limits to the hardware itself. In the present instance the equation solution is 64 processors * 32 members of a plex * 900 MIPS per processor running in operating system instances that support 4 terabytes of memory per instance.

The economic equation is less straightforward. It sometimes pays to replace mainframes in the middle of their useful life, for instance, if the vendor announces the new system in the middle of the depreciation arc of your present-generation installation, it jiggles the depreciation and makes yer numbers bad, watch out for the bears of Wall Street!

Of course this power, and this fiduciary responsibility, only apply to Big Biz and Big Guv. The question that gets asked over and over again by PC/Workstation jockeys trying to understand mainframing is: Is mainframe usage utilitarian, is it economical?

And the answer is simple: yes, most emphatically, if you're moving that much data live. The differential between trying to support such tasks on networked clusters of workstation/server architecture in terms of personnel requirements, security and reliability (five nines, anyone?) and supporting them on a SysPlex [1 ,2 ] makes it a no-brain, hands-down win for the venerable mainframe approach, the veteran opines.