This is almost an impossibility using a multi-tasking OS on two separate machines.

Multi-tasking is non-determanistic, which is to say, that there is no trivial way to predict exactly when any given application on a multi-tasking machine will receive a timeslice. There are simply too many system processes competing for those timeslices, and no way to predict when any of those system processes will become ready for scheduling, nor how much of their timeslice they will require to complete their task.

Even simple variations, like one machine being configured as a server and the other as a workstation will mean that both machines are probably running different suites of system applications--and probably, different thread quantums (timeslices).

You might be able to get this level of synchronicity using a single tasking (DOS-like) OS if the hardware was identical on both machines, but even that is unlikely.

The only realistic hope would be to use a real-time OS (like RTOS or QNX).