As a rule human processes and technologies evolve in response to perceived problems and opportunities, but then freeze into place until some sufficiently powerful external force is applied to dislodge them.
Prior to WWI and the Spanish flu, for example, a lot of Europe's aristocracy (and military leaders) reacted to the evolution of the horseless carriage by having theirs drawn by horses - and COBOL, when adopted for digital data processing in 1958, both implemented the the 1930s machine instruction set and inherited its card processing model in which a prepended card set (now expressed as JCL) effectively created a virtual machine environment for each job.
By the late 1950s, however, it was becoming clear that CPU costs and performance were significantly outstripping memory development and therefore that a way had to be found to more effectively share the CPU between processes.
From that three solutions evolved:
- the do nothing solution saw existing processes continued into what is now zOS with machine partitioning used to allow everything (even apparently interactive jobs) to run as batch jobs in dedicated machine environments;
- treating card decks as unitary and interleaving them on expensive components like the CPU led to what is now zVM and machine virtualization; and,
- card level interleaving - breaking up jobs to schedule each step for separate execution in a shared memory environment - led to today's Unix scheduler and memory management technologies.
Absent compelling force people continue doing what they know so zOS still implements 1930s procedures, data processing people moving into Wintel management positions have brought virtualization and related ideas with them, and IBM recent defense of a press release claiming a $26 million z10 the equal of 1,500 dual core Sun Opterons costing a cumulative $1.8 million provided that 1,350 of them are idle hasn't moved the industry to gales of laughter.
But it should have - and sooner or later it will, because IBM's own numbers show beyond all reasonable doubt that the first two solutions: partitioning and virtualization, make about as much sense today as hauling cars around with horses.