On Sat, 29 Jul 2017, at 14:53, [email protected] wrote: > I always thought that CPUS (engines) were physical hardware. > Is this not true anymore ?
Even in the hardware, a cpu runs microcode, which is to say its instruction set isn't hardwired into it, but programmed. But I doubt that's what you mean. Even as far back as the 1988, when IBM introduced PR/SM, it began to get complicated. https://en.wikipedia.org/wiki/PR/SM Different LPAR (logical partitions) could share physical resources, eg have 30% of the use of 2 cpus in a box, plus optionally some more. So for example on a machine with 4 cpus you could simultanously have more than 4 MVS images running at once, and each of them could think they had more than one cpu available. It was possible (though I never did so, and I'm not sure if the site where I worked ever did so) to have partitions running using minimal resources and then when some need arose, for the proportion of overall resources to change dynamically between partitions. Twenty years ago, sysplexes were getting going and I know IBM were aiming at sites being able to move workload (eg running IMS or CICS work) on one lpar within one machine and have it move all by itself to another machine in the sysplex either for planned migration of work from one machine to another or so that if the first machine failed the other could continue the work. I'm sure that's now considered routine. -- Jeremy Nicoll - my opinions are my own.
