Volume 2, Issue 2, February 2012 ISSN: 2277 128X

International Journal of Advanced Research in

Computer Science and Software Engineering
Research Paper
Available online at: www.ijarcsse.com

A Comparative Study of Software Reliability Models Using
SPC on Ungrouped Data

G.Krishna Mohan
Reader,
Dept. of Computer Science
P.B.Siddhartha college
Vijayawada.
km_mm_2000@yahoo.com

B. Srinivasa Rao
Associate Professor
Dept. of Compute Science
VRS & YRN College
Chirala
sreenibandla@yahoo.com

Dr. R.Satya Prasad
Associate Professor
Dept. of CS&E.
Acharya Nagrjuna University
Nagarjuna Nagar.
profrsp@gmail.com



AbstractControl charts are widely used for process monitoring. Software reliability process can be monitored efficiently by using
Statistical Process Control (SPC). It assists the software development team to identify failures and actions to be taken during software
failure process and hence, assures better software reliability. If not many, few researchers proposed SPC based software reliability
monitoring techniques to improve Software Reliability Process. In this paper we propose a control mechanism based on the
cumulative quantity between observations of time domain failure data using mean value function of Weibull and Goel-Okumoto
distribution, which are based on Non Homogenous Poisson Process (NHPP). The Maximum Likelihood Estimation (MLE) method is
used to derive the point estimators of the distributions.

Keywords Statistical Process Control, Software reliability, Weibull Distribution, Goel-Okumoto distribution, Mean Value function,
Probability limits, Control Charts.

I. INTRODUCTION
Software reliability assessment is important to evaluate and
predict the reliability and performance of software system,
since it is the main attribute of software. To identify and
eliminate human errors in software development process and
also to improve software reliability, the Statistical Process
Control concepts and methods are the best choice. SPC
concepts and methods are used to monitor the performance of
a software process over time in order to verify that the process
remains in the state of statistical control. It helps in finding
assignable causes, long term improvements in the software
process. Software quality and reliability can be achieved by
eliminating the causes or improving the software process or its
operating procedures [1].
The most popular technique for maintaining process control
is control charting. The control chart is one of the seven tools
for quality control. Software process control is used to secure
the quality of the final product which will conform to
predefined standards. In any process, regardless of how
carefully it is maintained, a certain amount of natural
variability will always exist. A process is said to be
statistically in-control when it operates with only chance
causes of variation. On the other hand, when assignable
causes are present, then we say that the process is statistically
out-of-control.
The control charts can be classified into several categories,
as per several distinct criteria. Depending on the number of
quality characteristics under investigation, charts can be
divided into univariate control charts and multivariate control
charts. Furthermore, the quality characteristic of interest may
be a continuous random variable or alternatively a discrete
attribute. Control charts should be capable to create an alarm
when a shift in the level of one or more parameters of the
underlying distribution or a non-random behavior occurs.
Normally, such a situation will be reflected in the control
chart by points plotted outside the control limits or by the
presence of specific patterns. The most common non-random
patterns are cycles, trends, mixtures and stratification [2]. For
a process to be in control the control chart should not have any
trend or nonrandom pattern.
SPC is a powerful tool to optimize the amount of
information needed for use in making management decisions.
Statistical techniques provide an understanding of the business
Volume 2, issue 2, February 2012 www.ijarcsse.com
2012, IJARCSSE All Rights Reserved

baselines, insights for process improvements, communication
of value and results of processes, and active and visible
involvement. SPC provides real time analysis to establish
controllable process baselines; learn, set, and dynamically
improves process capabilities; and focus business areas which
need improvement. The early detection of software failures
will improve the software reliability. The selection of proper
SPC charts is essential to effective statistical process control
implementation and use. The SPC chart selection is based on
data, situation and need [3]. Many factors influence the
process, resulting in variability. The causes of process
variability can be broadly classified into two categories, viz.,
assignable causes and chance causes.
The control limits can then be utilized to monitor the failure
times of components. After each failure, the time can be
plotted on the chart. If the plotted point falls between the
calculated control limits, it indicates that the process is in the
state of statistical control and no action is warranted. If the
point falls above the UCL, it indicates that the process average,
or the failure occurrence rate, may have decreased which
results in an increase in the time between failures. This is an
important indication of possible process improvement. If this
happens, the management should look for possible causes for
this improvement and if the causes are discovered then action
should be taken to maintain them. If the plotted point falls
below the LCL, It indicates that the process average, or the
failure occurrence rate, may have increased which results in a
decrease in the failure time. This means that process may have
deteriorated and thus actions should be taken to identify and
the causes may be removed. It can be noted here that the
parameter a, b should normally be estimated with the data
from the failure process. Since a, b are the parameters in the
proposed distributions, any traditional estimator can be used.
The control limits for the chart are defined in such a
manner that the process is considered to be out of control
when the time to observe exactly one failure is less than LCL
or greater than UCL. Our aim is to monitor the failure process
and detect any change of the intensity parameter. When the
process is normal, there is a chance for this to happen and it is
commonly known as false alarm. The traditional false alarm
probability is to set to be 0.27% although any other false
alarm probability can be used. The actual acceptable false
alarm probability should in fact depend on the actual product
or process [9].
II. LITERATURE SURVEY
This section presents the theory that underlies the proposed
distributions and maximum likelihood estimation for complete
data. If t is a continuous random variable with
pdf:
1 2
( ; , , , )
k
f t u u u . Where
1 2
, , ,
k
u u u are k unknown
constant parameters which need to be estimated, and cdf:
( ) F t . Where, The mathematical relationship between the
pdf and cdf is given by:
( ) ( )
( )
d F t
f t
dt
= . Let a denote the
expected number of faults that would be detected given
infinite testing time in case of finite failure NHPP models.
Then, the mean value function of the finite failure NHPP
models can be written as: ( ) ( ) m t aF t = . where, F(t) is a
cumulative distribution function. The failure intensity function
( ) t in case of the finite failure NHPP models is given by:
( ) '( ) t aF t = [8].
A. NHPP model
The Non-Homogenous Poisson Process (NHPP) based
software reliability growth models (SRGMs) are proved to be
quite successful in practical software reliability engineering
[4]. The main issue in the NHPP model is to determine an
appropriate mean value function to denote the expected
number of failures experienced up to a certain time point.
Model parameters can be estimated by using Maximum
Likelihood Estimate (MLE). Various NHPP SRGMs have
been built upon various assumptions. Many of the SRGMs
assume that each time a failure occurs, the fault that caused it
can be immediately removed and no new faults are introduced.
Which is usually called perfect debugging. Imperfect
debugging models have proposed a relaxation of the above
assumption [5,6].
Let ( ) { }
, 0 N t t > be the cumulative number of software
failures by time t. m(t) is the mean value function,
representing the expected number of software failures by time
t. ( ) t is the failure intensity function, which is
proportional to the residual fault content. Thus
( ) (1 )
bt
m t a e

= and
( )
( )
( ( ))
dm t
t b a m t
dt
= =
.
where a denotes the initial number of faults contained in a
program and b represents the fault detection rate. In software
reliability, the initial number of faults and the fault detection
rate are always unknown. The maximum likelihood technique
can be used to evaluate the unknown parameters. In NHPP
SRGM ( ) t can be expressed in a more general way as
( ) ( ) ( ) ( )
( ) dm t
t b t a t m t
dt
= = (

. where ( ) a t is the time-
dependent fault content function which includes the initial and
introduced faults in the program and ( ) b t is the time-
dependent fault detection rate. A constant ( ) a t implies the
perfect debugging assumption, i.e no new faults are
introduced during the debugging process. A constant ( ) b t
implies the imperfect debugging assumption, i.e when the
faults are removed, then there is a possibility to introduce new
faults.
B. Goel-Okumoto distribution
The Goel-Okumoto model is a simple NonHomogenous
Poisson Process (NHPP) model with the mean value function
( ) ( ) 1
bt
m t a e

=
[12]. Where the parameter a is the
number of initial faults in the software and the parameter b
is the fault detection rate. The corresponding failure intensity
Volume 2, issue 2, February 2012 www.ijarcsse.com
2012, IJARCSSE All Rights Reserved

function is given by
( )
bt
t abe

=
. The probability density
function of a Goel-Okumoto model has the form:
( )
bt
f t be

= . The corresponding cumulative distribution

function is: ( ) 1
bt
F t e

= .
C. Weibull distribution
The Weibull distribution is a generalization of exponential
distribution, which is recovered for = 1. Although the
exponential distribution has been widely used for times-
between-event, Weibull distribution is more suitable as it is
more flexible and is able to deal with different types of aging
phenomenon in reliability. Hence in reliability monitoring of
equipment failures, the Weibull distribution is a good
alternative. The probability density function of a two-
parameter Weibull model has the form:
( )
( )
1
( )
bt
f t b bt e
|
|
|


= . Where b > 0 is a scale parameter
and 0 | > is a shape parameter. The corresponding
cumulative distribution function is: ( )
( )
1
bt
F t e
|

= . The
mean value function
( )
( ) 1
n
bt
m t a e
|

(
=
(

. The failure
intensity function is given as:
1 ( )
( ) .
bt
t ab t e
|
| |
|

= .
D. MLE (Maximum Likelihood) Parameter Estimation
The idea behind maximum likelihood parameter estimation
is to determine the parameters that maximize the probability
(likelihood) of the sample data. The method of maximum
likelihood is considered to be more robust (with some
exceptions) and yields estimators with good statistical
properties. In other words, MLE methods are versatile and
apply to many models and to different types of data. Although
the methodology for maximum likelihood estimation is simple,
the implementation is mathematically intense. Using today's
computer power, however, mathematical complexity is not a
big obstacle. If we conduct an experiment and obtain N
independent observations,
1 2
, , ,
N
t t t . The likelihood
function [7] may be given by the following product:
( )
1 2 1 2 1 2
1
, , , | , , , ( ; , , , )
N
N k i k
i
L t t t f t u u u u u u
=
=
[

Likely hood function by using (t) is:
1
( )
n
i
i
L t
=
=
[

The logarithmic likelihood function is given by:

| |
1
1
log log ( )
log ( ) ( )
n
i
i
n
i n
i
L t
t m t

=
=
| |
=
|
\ .
=
[

The maximum likelihood estimators (MLE) of
1 2
, , ,
k
u u u are obtained by maximizing L or A, where Ais
ln L . By maximizing , which is much easier to work with
than L, the maximum likelihood estimators (MLE) of
1 2
, , ,
k
u u u are the simultaneous solutions of k equations
such as:
( )
0
j
u
c A
=
c
, j=1,2,,k
The parameters a and b are estimated as follows. The
parameter b is estimated by iterative Newton Raphson
Method using
1
( )
'( )
n
n n
n
g b
b b
g b
+
=
, which is substituted in
finding a.
III. ILLUSTRATING THE MLE METHOD
A. Goel-Okumoto parameter estimation
The likelihood function is given as, ( )
1
N
bt
i
L abe

=
=
[

Taking the natural logarithm on both sides, The Log
Likelihood function is given as:
( ) ( )
1
log log( ) [1 ]
i n
n
bt bt
i
L abe a e

=
=

.
Taking the Partial derivative with respect to a and
equating to 0. (i.e
log
0
L
a
c
=
c
).
( )
1
n
bt
n
a
e

=
(

Taking the Partial derivative with respect to b and
equating to0.(i.e
log
( ) 0
L
g b
b
c
= =
c
).
( )
( )
( )
1
( ) 0
1
n
n
bt
n
i n
bt
i
n e
g b t nt
b
e

=
= + =

Taking the partial derivative again with respect to b and
equating to 0. (i.e
2
2
log
'( ) 0
L
g b
b
c
= =
c
).
( )
( )
( )
( )
( )
2
2 2
1
'( )
1
1
n
n
n
n
bt
bt
n
bt
bt
n e
g b nt e
b
e
e

= +
`

)

B. Weibull parameter estimation
The likelihood function, assuming 2 | = (Rayleigh) is
given as,
2
2 ( )
1
2 .
N
bt
i
L ab t e

=
=
[

Taking the natural logarithm on both sides, The Log
Likelihood function is given as:
2 2
( ) 2 ( )
1
log log(2 ) [1 ]
n
n
bt bt
i
i
L ab t e a e

=
=

.
Taking the Partial derivative with respect to a and
equating to 0. (i.e
log
0
L
a
c
=
c
).
( )
2
1
n
bt
n
a
e

=
(

(


Taking the Partial derivative with respect to b and
equating to0.(i.e
log
( ) 0
L
g b
b
c
= =
c
).
( )
( )
( )
2
2
2
2
1
2. . . . 2
( ) 2 0
1
n
n
bt
n
n
i
bt
i
nbt e n
g b b t
b
e

=
= =

Volume 2, issue 2, February 2012 www.ijarcsse.com
2012, IJARCSSE All Rights Reserved

Taking the partial derivative again with respect to b and
equating to 0. (i.e
2
2
log
'( ) 0
L
g b
b
c
= =
c
).
( )
( )
( )
( )
( )
( )
2 2
2
2
2 2
2 2
2 2
1
2 . 1
'( ) 2 2 2
1
1
n n
n
n
bt bt
n
n
i n
bt
bt
i
b t e e
g b n t nt
b
e
e

=


| |
=
`
|
\ .

C. Distribution of Time between failures
Based on the inter failure data given in Table 1, we
compute the software failures process through Mean Value
Control chart. We used cumulative time between failures data
for software reliability monitoring using Goel-Okumoto and
Weibull distributions. The use of cumulative quality is a
different and new approach, which is of particular advantage
in reliability.
a
.
and b
.
are Maximum Likely hood Estimates of
parameters and the values can be computed using iterative
method for the given cumulative time between failures data
[10] shown in table 1. Using a and b values we can
compute ( ) m t .
TABLE 1. TIME BETWEEN FAILURES OF A SOFTWARE
Failure
Number
Time
between
failure(h)
Failure
Number
Time
between
failure(h)
1 30.02 16 15.53
2 1.44 17 25.72
3 22.47 18 2.79
4 1.36 19 1.92
5 3.43 20 4.13
6 13.2 21 70.47
7 5.15 22 17.07
8 3.83 23 3.99
9 21 24 176.06
10 12.97 25 81.07
11 0.47 26 2.27
12 6.23 27 15.63
13 3.39 28 120.78
14 9.11 29 30.81
15 2.18 30 34.19


Assuming an acceptable probability of false alarm of 0.27%,
the control limits can be obtained as [10]:

( )
1 0.99865
bt
U
T e
|

= =

( )
1 0.5
bt
C
T e
|

= =

( )
1 0.00135
bt
L
T e
|

= =

These limits are converted to ( )
U
m t , ( )
C
m t and ( )
L
m t
form. They are used to find whether the software process is in
control or not by placing the points in Mean value chart
shown in figure 1 and figure 2. A point below the control limit
( )
L
m t indicates an alarming signal. A point above the
control limit ( )
U
m t indicates better quality. If the points are
falling within the control limits, it indicates the software
process is in stable condition [11]. The values of parameter
estimates and the control limits are given in table 2 and 3
respectively.
TABLE 2. PARAMETER ESTIMATES
model a b
GO 31.698171 0.003962
Weibull 30.051592 0.003416

TABLE 3. CONTROL LIMITS.
model
) (
U
t m

) (
C
t m

) (
L
t m

GO 31.676760 21.132114 0.085469
Weibull 30.011170 15.025870 0.040570
TABLE 4. MEAN SUCCESSIVE DIFFERENCES OF GO
FN m(t) SD
1 3.554578 0.160101
2 3.714687 2.383587
3 6.098274 0.137569
4 6.235844 0.343684
5 6.579527 1.279946
6 7.859432 0.481484
7 8.340916 0.351758
8 8.692674 1.836638
9 10.529312 1.060330
10 11.589642 0.037410
11 11.627052 0.489356
12 12.116408 0.261248
13 12.377656 0.684916
14 13.062573 0.160266
15 13.222838 1.102518
16 14.325356 1.683122
17 16.008478 0.172479
18 13.180956 0.117592
19 16.298549 0.249935
20 16.548483 3.690661
21 20.239144 0.749363
22 20.988508 0.167971
23 21.156479 5.293999
24 26.450479 1.441653
25 27.892132 0.034077
26 27.926209 0.226497
27 28.152706 1.348363
28 29.501069 0.252475
29 29.753545 0.246358
30 29.999903
TABLE 5. MEAN SUCCESSIVE DIFFERENCES OF WEIBULL
FN m(t) SD
1 0.314371 0.030704
2 0.345076 0.657725
3 1.002801 0.050307
4 1.053108 0.132025
5 1.185134 0.575065
6 1.760199 0.252180
7 2.012380 0.197261
8 2.209641 1.219663
Volume 2, issue 2, February 2012 www.ijarcsse.com
2012, IJARCSSE All Rights Reserved

9 3.429305 0.859242
10 4.288547 0.032507
11 4.321054 0.439360
12 4.760415 0.245447
13 5.005863 0.680255
14 5.686118 0.166975
15 5.853094 1.230688
16 7.083782 2.161267
17 9.245050 0.240957
18 9.486008 0.166350
19 9.652358 0.359124
20 10.011482 6.120127
21 16.131610 1.396357
22 17.527968 0.317624
23 17.845592 9.491850
24 27.337443 1.649185
25 28.986628 0.029822
26 29.016451 0.186649
27 29.203101 0.697874
28 29.900976 0.058849
29 29.959825 0.040168
30 29.999994


Figure 1 and 2 are obtained by placing the time between
failures cumulative data shown in tables 3, 4 on y axis and
failure number on x axis, and the values of control limits are
placed on Mean Value chart. The Mean Value chart of Goel-
Okumoto shows that the 10
th
and 25
th
failure data has fallen
below ( )
L
m t . The Mean Value chart of weibull shows that
the 1
st
,10
th
and 25
th
failure data has fallen below ( )
L
m t . The
successive differences of mean values below ( )
L
m t indicates
the failure process. In the present scenario, It is significantly
early detection of failure through weibull using Mean Value
Chart. The software quality is determined by detecting failures
at an early stage. The Remaining Failure data shown in figure
1 are in stable condition. No failure data fall outside
the ( )
U
m t . It does not indicate any alarm signal.



Failure Control Chart
UCL=31.676760
CL=21.132114
LCL=0.085469
0.010000
0.100000
1.000000
10.000000
100.000000
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Failure Number
M
e
a
n

V
a
l
u
e

S
u
c
c
e
s
s
i
v
e

D
i
f
f
e
r
e
n
c
e
s

Figure: 1 GO Failure Control Chart
Mean Value Chart
UCL=30.011022
CL=15.025796
LCL=0.040570
0.010000
0.100000
1.000000
10.000000
100.000000
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
Failure Number
M
e
a
n

V
a
l
u
e

S
u
c
c
e
s
s
i
v
e

D
i
f
f
e
r
e
n
c
e
s

Figure: 2 Weibull Mean Value Chart
IV. CONCLUSION
The given 30 inter failure times are plotted through the
estimated mean value function against the failure serial order.
The parameter estimation is carried out by Newton Raphson
Iterative method for the models. The graphs have shown out
of control signals i.e below the LCL. Hence we conclude that
our method of estimation and the control chart are giving a
+ve recommendation for their use in finding out preferable
control process or desirable out of control signal. By
observing the Mean value Control chart we identified that the
failure situation is detected at 10th and 25th point of table-4,
1st ,10th, 25th and 29th point of table-5 i.e ailure data has
fallen below ( )
L
m t . The successive difference of mean
values below ( )
L
m t indicates the failure process. In the
present scenario, It is significantly early detection of failure
through weibull using Mean Value Chart. The software
quality is determined by detecting failures at an early stage for
the corresponding ( ) m t , which is below ( )
L
m t . It indicates
that the failure process is detected at an early stage compared
with Xie et. a1 (2002) control chart [10], which detects the
failure at 23rd point for the inter failure data above the UCL.
Hence our proposed Mean Value Chart detects out of control
situation at an earlier than the situation in the time control
chart. The early detection of software failure will improve the
software Reliability. When the time between failures is less
than LCL, it is likely that there are assignable causes leading
to significant process deterioration and it should be
investigated. On the other hand, when the time between
failures has exceeded the UCL, there are probably reasons that
have lead to significant improvement.
REFERENCES
[1] Kimura, M., Yamada, S., Osaki, S., 1995. Statistical Software
reliability prediction and its applicability based on mean time between
failures. Mathematical and Computer Modeling Volume 22, Issues
10-12, Pages 149-155.
[2] Koutras, M.V., Bersimis, S., Maravelakis,P.E., 2007. Statistical
process control using shewart control charts with supplementary Runs
rules Springer Science + Business media 9:207-224.
[3] MacGregor, J.F., Kourti, T., 1995. Statistical process control of
multivariate processes. Control Engineering Practice Volume 3, Issue
3, March 1995, Pages 403-414 .
[4] Musa, J.D., Iannino, A., Okumoto, k., 1987. Software Reliability:
Measurement Prediction Application. McGraw-Hill, New York.
[5] Ohba, M., 1984. Software reliability analysis model. IBM J. Res.
Develop. 28, 428-443.
Volume 2, issue 2, February 2012 www.ijarcsse.com
2012, IJARCSSE All Rights Reserved

[6] Pham. H., 1993. Software reliability assessment: Imperfect debugging
and multiple failure types in software development. EG&G-RAAM-
10737; Idaho National Engineering Laboratory.
[7] Pham. H., 2003. Handbook Of Reliability Engineering, Springer.
[8] Pham. H., 2006. System software reliability, Springer.
[9] Swapna S. Gokhale and Kishore S.Trivedi, 1998. Log-Logistic
Software Reliability Growth Model. The 3rd IEEE International
Symposium on High-Assurance Systems Engineering. IEEE Computer
Society.
[10] Xie, M., Goh. T.N., Ranjan.P., Some effective control chart
procedures for reliability monitoring -Reliability engineering and
System Safety 77 143 -150 2002.
[11] Satya Prasad, R., Half logistic distribution for software reliability
growth model, Ph.D thesis, 2007.
Goel, A.L., Okumoto, K., 1979. Time-dependent errordetection rate model for
software reliability and other performance measures. IEEE Trans. Reliab. R-
28, 206-211.
AUTHOR PROFILE:
FIRST AUTHOR:
G. Krishna Mohan is working as a Reader in
the Department of Computer Science,
P.B.Siddhartha College, Vijayawada. He
obtained his M.C.A degree from Acharya
Nagarjuna University in 2000, M.Tech from
JNTU, Kakinada, M.Phil from Madurai
Kamaraj University and pursuing Ph.D at
Acharya Nagarjuna University. His research interests lies in
Data Mining and Software Engineering. He has published
several research papers in National and International Journals.
SECOND AUTHOR:
B. Srinivasa rao received the Master Degree in
Computer Science and Engineering from Dr
MGR Deemed University, Chennai, Tamil
Nadu, India. He is Currently working as
Associate Professor in PG Department of
Computer Applications, VRS & YRN
College, Chirala, Andhra Pradesh, India. His research interests
include software reliability, Cryptography and Computer
Networks. He has published several papers in National and
International Journals.










THIRD AUTHOR:
Dr. R. Satya Prasad received Ph.D. degree in
Computer Science in the faculty of
Engineering in 2007 from Acharya Nagarjuna
University, Andhra Pradesh. He received gold
medal from Acharya Nagarjuna University for
his outstanding performance in Masters
Degree. He is currently working as Associate
Professor and H.O.D, in the Department of Computer Science
& Engineering, Acharya Nagarjuna University. His current
research is focused on Software Engineering. He has published
several papers in National & International Journals.