Applying Software Failure Modes and Effects Analysis to Interfaces

Software failure modes and effects analysis (SFMEA) is sometimes applied to new mission-critical and safety-critical system development. This kind of analysis, like its older cousin Hardware FMEA (HFMEA), tries to determine all possible types of failure for each component, one by one, and attempts to predict system-level effects for each failure. While software variables and classes don't fail in the sense that hardware fails, variables do sometimes assume unexpected values [1] and class methods don't always perform as expected. Every software system includes software and hardware interfaces. In large system developments, different groups of people usually develop different parts of the software, with developers depending on interface requirement specifications (IRSs) to guide their design. Unfortunately, IRSs are sometimes like brick walls that separate developers. Since software designers are naturally most concerned with their own designs rather than those on the other side of the wall, there is tremendous potential for unidentified failures across the interface. Applying an SFMEA to both sides of the wall is an essential first step for understanding consequences of failures, but an analysis that fails to rigorously analyze the interface can lead to conclusions that are both incomplete and incorrect. Fortunately, SFMEA techniques can be expanded to include considerations that apply to any kind of interface. The idea is to apply a step-by-step analysis sequence to determine what could go wrong at an interface and the subsequent effects on the system software. SFMEA that includes thorough interface analyses provides a more complete picture of system robustness.