Architecture and Real-Time Systems (ARTS) Laboratory

Currently Available Reliability-Measurement Tools

Traditional Measures

These include reliability, throughput, availability, and measures derived from these (e.g., point availiability, capacity). These are difficult to use for two reasons:

• They do not take key dynamic factors into account (e.g., scheduling, recovery)
• They are computer-centric

Application-Specific Measures

These include:

• Performability (Meyer, 1980) - accomplishment levels of the application must be precisely defined and the computer performance linked to them.
• Cost Functions and Probability of Dynamic Failure (Krishna & Shin, 1983) - links the computer response time to the performance of the application
• These are measured by tools including HARP, SHARPE, ULTRASAN, CARE-III, etc.

Such measures are often difficult to use for real-time applications because they require complete details about the application, which can be unavailable.