Modeling impairments of computing systems considering hardware–software interaction failures and dependability quantification

The computing systems deployed in safety-critical applications often comprise complex hardware with highly intensive software/firmware. The development and validation of computing systems for safety-critical applications shall comply with the Design Assurance Level-A as per RTCA DO-254/DO-178 or Safety Integrity Level-4 as per IEC-61508 towards certification. However, adherence to this process does not assure dependability. Quantifying the reliability to assess the risk associated with using these systems for safety-critical applications is necessary. Hence, dependability quantification of the computing system has been undertaken in this research, with significant improvements over the conventional approaches. In the classical approach, hardware faults and software errors were treated as independent events, and failures arising from interactions were ignored. It is proposed to model dependent states arising due to hardware–software interaction, such as hardware-triggered software failure and software-triggered hardware failures, in addition to hardware and software failures, to model the failure characteristics of the system completely. A stochastic Petri Net (SPN) based methodology is adopted to model the error/fault propagation by considering all possible places, transitions, and tokens. SPN is then transformed into a Continuous-Time Markov Chain to quantify reliability analytically. This enhanced methodology enables more accurate dependability quantification and risk assessment.