Information Technology Reference
In-Depth Information
developed at Purdue University. Using Mothora, the tester can create and execute
test cases, measure test case adequacy, determine input-output transfer function
correctness, locate and remove faults or bugs, and control and document the test.
For run-time checking and debugging aids, you can use NuMega's Boundschecker
6
or Rational's Purify.
7
Both can both check and protect against memory leaks and
pointer problems. Ballista COTS Software Robustness Testing Harness
8
is a full-
scale automated robustness testing tool. The first version supports testing up to
233 POSIX
9
function calls in UNIX operating systems. The second version also
supports testing of user functions provided that the data types are recognized by the
testing server. The Ballista testing harness gives quantitative measures of robustness
comparisons across operating systems. The goal is to test automatically and to harden
commercial off-the-shelf (COTS) software against robustness failures.
In experimental design, decision variables are referred to as
factors
and the output
measures are referred to as
response, software metric
(e.g., complexity), or
functional
requirement
(e.g., GUI). Factors often are classified into
control
and
noise
factors.
Control factors are within the control of the design team, whereas noise factors are
imposed by operating conditions and other internal or external uncontrollable factors.
The objective of software experiments usually is to determine settings to the software
control factors so that software response is optimized and system random (noise)
factors have the least impact on system response. You will read more about the
setup and analysis of designed experiments in the following chapters.
6.5 A NOTE ON NORMAL DISTRIBUTION AND
NORMALITY ASSUMPTION
Normal distribution is used in different domains of knowledge, and as such, it is
standardized to avoid the taxing effort of generating specialized statistical tables.
A standard normal has a mean of 0 and a standard deviation of 1, and functional
requirement,
y
, values are converted into
Z
-scores or Sigma levels using
Z
i
=
(
y
i
−
µ
)
σ
transformation. A property of the normal distribution is that 68% of all of its ob-
servations fall within a range of
±
1 standard deviation from the mean, and a range
of
2 standard deviations includes 95% of the scores. In other words, in a normal
distribution, observations that have a
Z
-score (or Sigma value) of less than
±
−
2or
more than
2 have a relative frequency of 5% or less. A
Z
-core value means that a
value is expressed in terms of its difference from the mean, divided by the standard
deviation. If you have access to statistical software such as Minitab, you can explore
the exact values of probability associated with different values in the normal distri-
bution using the Probability Calculator tool; for example, if you enter the
Z
value
(i.e., standardized value) of 4, the associated probability computed will be less than
+
6
http://www.numega.com/devcenter/bc.shtml.
7
http://www.rational.com/products/purify unix/index.jtmpl.
8
http://www.cs.cmu.edu/afs/cs/project/edrc-ballista/www/.
9
POSIX (pronounced/pvziks/) or “Portable Operating System Interface [for Unix]”.
Search WWH ::
Custom Search