nii shonan-meeting-gsrm-20141021
DESCRIPTION
“Predicting Release Time Based on Generalized Software Reliability Model”, by Hironori Washizaki COMPUTATIONAL INTELLIGENCE FOR SOFTWARE ENGINEERING NII Shonan Meeting, Oct 21, 2014 Development environments have changed drastically, development periods are shorter than ever and the number of team members has increased. These changes have led to difficulties in controlling the development activities and predicting when the development will end. We propose a generalized software reliability model (GSRM) based on a stochastic process, and simulate developments that include uncertainties and dynamics. We also compare our simulation results to those of other software reliability models. Using the values of uncertainties and dynamics obtained from GSRM, we can evaluate the developments in a quantitative manner. Additionally, we use equations to define the uncertainty regarding the time required to complete a development, and predict whether or not a development will be completed on time.TRANSCRIPT
Generalized Software Reliability Model
(GSRM)Hironori Washizaki
Waseda University
National Institute of InformaticsTwitter: @Hiro_Washi [email protected]
http://www.washi.cs.waseda.ac.jp/
Kiyoshi Honda, Hironori Washizaki, YoshiakiFukazawa, “A Generalized Software Reliability Model Considering Uncertainty and Dynamics in Development,” PROFES 2013
NII Shonan Meeting, Oct 21, 2014
Motivation
• When can we release software?• How many efforts are necessary for further
testing?
2Time
Software Reliability Model (SRM)
3
Counting the defects a day or a week
Approximate actual data to a curve and predict defects
At this time, 95% of all defects will be found
Prediction model
Types of SRM
• Statistic analysis model– From actual data approximate
to a curve.– Gompertz model– Logistic model
• Stochastic process model– The detection of defects
follows stochastic process – Non-homogeneous Poisson
process(NHPP) model [Goel]4[Goel] A.L. Goel and K. Okumoto A non-homogeneous poisson process model for
software reliability and other performance measures, 1979
Further Challenges in SRM
• Uncertainty– Actual projects have many uncertain elements
which cause defects.– E.g. changes of specifications
• Dynamicity– Actual projects have some time dependency.– E.g. changes of developers
6
Idea: Generalized SRM• Conventional Logistic Model
• Assumptions– Number of defects that can be found is variable depending
on time.– Number of defects that can be found contains uncertainty,
which can be simulated with Gaussian white noise.
7
Dynamicity Uncertainty
A generalized software reliability model considering uncertainty and dynamics in development. PROFES ’13
Uncertainty
8
the uncertainty is greater at the start of the project than at the end.
The uncertainty is constant at any given time.
The uncertainty increases near the end.
Dynamicity
9
The number of developers is constant.
The number of developersper unit time changes at a certain time.
The number of developers per unit time increase near the end.
Case (Industry)
13
abyss_data-api abyss_data-api abyss_data-api
Module Current PredictedTotal
PredictedCurrent
PredictedEnd Day
XYZ 147 144 134 156
Almost all defects seem to be detected.Now more debugging rather than testing.
#Defects #Predicted Total DefectsPredication Error
Case (OSS)
14Kiyoshi Honda, Hironori Washizaki, Yoshiaki Fukazawa, “Predicting the Release Time Based on a Generalized Software Reliability Model (GSRM),” COMPSAC’14