It has been documented that microprocessor performance doubles about every 21 months. Much is published and reported on the technology that delivers this impressive computing power. Much less is said, however, about the unique Input/Output (I/O) demands that are presented when using these microprocessors in high-performance, real-time, multiprocessing environments. Raw computing power is seldom questioned anymore. Of more concern today is the ability of a computer system to deliver data to and from these high-performance processors.

For example, it is not difficult to select a computing engine that is capable of performing the computations necessary to drive a Full Flight Simulator (FFS) or a Weapons Systems Trainer (WST). It is, however, a significantly greater challenge to determine how the simulation I/O can be performed so as to eliminate bottlenecks and latencies. The training value of a simulator can easily be lowered by the stepping or jumping of an instrument, visual system, or motion base that is due to the inability of the I/O to keep up with the processors.

This paper will explore some of the technology available that can be used to "feed" today's high-performance, real-time, multiprocessing systems. Both advances in hardware and software will be discussed, advances that give developers the tools they need to deliver I/O to and from a simulator with determinism and realism.