I haven't seen any references here to what I believe was the "original" implementation of virtual machines -- a system that came out of IBM's Cambridge Research Center, and was eventually marketed for the 360/67 mainframe in the early 1970s as CP-67.
Rather than running user programs, CP-67 multitasked operating systems. Each client system ran believing that it was operating in supervisor mode on a bare 360/65. In fact, the clients were all running in user mode. CP-67 trapped the privileged instructions (including all i/o instructions), which it emulated on behalf its clients; and it used the Model 67's Dynamic Address Translation unit (what we'd call a VMM today) to present a flat model of the 360's address space to each client operating system.
CP-67 was neither a bytecode interpreter nor a full-fledged emulator. All programs were compiled into 360 machine code, and most instructions (all of the non-privileged ones) executed natively. IBM referred to the client environment as a "virtual machine." The flat address space presented to each client (which could magically exceed the amount of physical memory in the machine) they called "virtual memory."
If you haven't already read it, I recommend the book Peopleware by Tom DeMarco and Timothy Lister. In addition to being highly readable, it offers some genuine insights into what makes Engineers productive.
Slashdot Mirror
Rather than running user programs, CP-67 multitasked operating systems. Each client system ran believing that it was operating in supervisor mode on a bare 360/65. In fact, the clients were all running in user mode. CP-67 trapped the privileged instructions (including all i/o instructions), which it emulated on behalf its clients; and it used the Model 67's Dynamic Address Translation unit (what we'd call a VMM today) to present a flat model of the 360's address space to each client operating system.
CP-67 was neither a bytecode interpreter nor a full-fledged emulator. All programs were compiled into 360 machine code, and most instructions (all of the non-privileged ones) executed natively. IBM referred to the client environment as a "virtual machine." The flat address space presented to each client (which could magically exceed the amount of physical memory in the machine) they called "virtual memory."
If you haven't already read it, I recommend the book Peopleware by Tom DeMarco and Timothy Lister. In addition to being highly readable, it offers some genuine insights into what makes Engineers productive.