SOFTWARE DESIGN FOR SIX SIGMA (DFSS) RISK MANAGEMENT PROCESS - Software Design for Six-Sigma: A Roadmap for Excellence - page 396

Software Failure Mode and Effects Analysis (SFMEA) 2

15.3.3

For a long time, the measured failure rate has been the standard for software quality
and, hence, reliability. In current computing environments and in the DFSS era, it
can be lagging, inapplicable, and even misleading. Consider server software with an
expanding number of clients. More users are likely to cause an increase in the failure
rate, though the software is not changed. Another example is software controlling a
machine tool. The machine tool is aging, causing more exception conditions to be
encountered by the program and, hence, more failures. The machine shop supervisor
sees a higher failure rate, even though the software remains the same.
Because there are problems with using failure rate as an indicator of quality in
existing software, we looked for alternatives for predicting software quality during
development that would continue to be valid in operation. The severity of failure
effects needed to be taken into account so that preventive DFSS actions could focus
on avoidance of the most severe failures. This latter requirement suggested a look
at software risk management, including tools such as FMEA. But although FMEA
for hardware is used widely (Yang & El-Haik, 2008), it rarely is encountered for
software. An obvious reason is that hardware generally is made up of parts with well-
known failure modes; there is no equivalent of this in software. Instead, software is
analyzed by “functions.” But these are subjective partitions, and there is usually no
certainty that all functions that can contribute to failure have been included.
FMEA 3 is a systematic method used to analyze products and processes by quali-
tatively determining their failure modes, causes of failure, and potential effects then
quantitatively classifying their risk estimate to prioritize better corrective and pre-
ventive actions and risk reduction measures required by the analysis.
Software FMEA (SFMEA) determines software effects of each failure mode of
each code component, one by one, identifies failures leading to specific end events,
has rules that differ from hardware analysis rules and is complex, its effects dependent
on time and state.
When SFMEA is extended further by a criticality analysis, the resulting technique
then is called failure mode and effects criticality analysis (FMECA). Failure mode and
effects analysis has gained wide acceptance by most industries. In fact, the technique
has adapted itself in many other forms such as concept FMEA, robust design FMEA, 4
process (manufacturing and service) FMEA, and use FMEA.
What makes SFMEA different from other applications?

Extended effects: Variables can be read and set in multiple places
Failure mode applicability: There can be different failure modes in different
places
Time dependency: Validity can depend on what is happening around it

2 See Chapter 16 for more details.
3 FMEAs were formally introduced in the late 1940s with the introduction of the United States Military
Procedure MIL-P-1629.
4 See Mekki (2006). See also Yang and El-Haik (2008) and El-Haik and Roy (2005).