SOFTWARE DESIGN FOR X - Software Design for Six-Sigma: A Roadmap for Excellence - page 375

TABLE 14.4 Defect Removal Techniques Efficiency

Defect Removal Technique Efficiency Range

Design Inspection 45%-60%

Code Inspection 45%-60%

Unit Testing 15%-45%

Regression test 15%-30%

Integration Test 25%-40%

Performance Test 20%-40%

System Testing 25%-55%

Acceptance Test 25%-35%

its intended function at time equals zero, reliability can be defined as the probability
that the software will continue to perform its intended function without failure for a
specified period of time under stated conditions.
Though the software has a reliable design, it is effectively unreliable when fielded,
which is actually the result of a substandard development process. Evaluating and
finding ways to attain high reliability are all aspects of software reliability.
The best option for software design for reliability is to optimize the returns from
software development “best practices.” Table 14.4 14

shows the difference in defect

removal efficiency between inspections and testing.
Most commercial companies do not measure defect removal in pretesting phases.
This leads to inspections that provide very few benefits. Unstructured inspections
result in weak results. Software belts simply do not know how to apply effectively
their efforts as reviewers to find defects that will lead to run-time failures. Inspection
results are increased by incorporating prevalent defect checklists based on historical
data and assigning reviewer perspectives to focus on vulnerable sections of designs
and code. By performing analysis techniques, such as failure analysis, static code
analysis, and maintenance reviews for coding standards compliance and complexity
assessments, code inspections become smaller in scope and uncover more defects.
Once inspection results are optimized, the combined defect removal results with
format testing and software quality assurance processes have the potential to remove
up to 99% of all inherent defects.
By redirecting their efforts upstream, most development organizations will see
greater improvements in software reliability with investments in design and code
inspections than further investments in testing (Table 14.5 15 ).
Software DFR practices increase confidence even before the software is executed.
The view of defect detection changes from relying solely on the test phases and
customer usage to one of phase containment across all development phases. By

14 See Silverman and De La Fuente, http://www.opsalacarte.com/pdfs/Tech Papers/Software Design for
Reliability - Paper.pdf.
15 See Silverman and De La Fuente, http://www.opsalacarte.com/pdfs/Tech Papers/Software Design for
Reliability - Paper.pdf.