PERFORMANCE MEASUREMENTS OF REAL-TIME COMPUTER SYSTEMS

Abstract

The performance of general purpose computers is typically measured in terms of Millions of Instructions per Second (MIPS) or Millions of Floating-Point Operations per Second (MFLOPS). Standard benchmark programs such as Whetstone, Dhrystone, and Linpack typically measure CPU speed in a single-task environment. However, a computer may have high CPU performance, but poor real-time capabilities. Therefore there is a need for performance measures specifically intended for real-time computer systems. This paper presents four methodologies, related metrics and benchmarks for objectively measuring real-time performance: (a) Tri-Dimensional Measure, (b) Process Dispatch Latency Time, (c) Rhealstone Metric, and (d) Vanada Benchmark. This proposed methodologies and related measures are applied in the performance evaluation of several real-time computer systems, and the results obtained are presented.