As another example of something orchestration-like ("run 17 processes where possible on this 100 node cluster"), there were also OpenMOSIX and Beowulf clusters on Linux, which if I remember right would load balance work across available systems, in the same way that the kernel's scheduler balances work across CPU cores.