the first attempt was on a multi-headed vax whose model number escapes me. it was pretty foul architecturally.
in early 1989 a restart was done for the 25-MHz mips CPU inside a 4-headed SGI shared-memory multiprocessor. the CPU had no synchronizing instructions but the SGI machine had a special piece of shared memory with locking properties. the kernel muxed locks onto sections of that memory. that system became real. a second machine became the file server. -rob
