Empirical Software Engineering

, Volume 15, Issue 3, pp 296–319 | Cite as

A multiplicative model of software defect repair times



We hypothesize that software defect repair times can be characterized by the Laplace Transform of the Lognormal (LTLN) distribution. This hypothesis is rooted in the observation that software defect repair times are influenced by the multiplicative interplay of several factors, and the lognormal distribution is a natural choice to model rates of occurrence of such phenomenon. Conversion of the lognormal rate distribution to an occurrence time distribution yields the LTLN. We analyzed a total of more than 10,000 software defect repair times collected over nine products at Cisco Systems to confirm our LTLN hypothesis. Our results also demonstrate that the LTLN distribution provides a statistically better fit to the observed repair times than either of the two most widely used repair time distributions, namely, the lognormal and the exponential. Moreover, we show that the repair times of subsets of defects, partitioned according to the Orthogonal Defect Classification (ODC) scheme also follow the LTLN distribution. Finally, we describe how the insights that lead to the LTLN repair time model allow us to consider and evaluate alternative process improvement strategies.


Software defect repair Lognormal Multiplicative effect 



We thank Cisco management, especially David Hsiao, for their support. The research at the Univ. of Connecticut was supported in part by a CAREER award from the National Science Foundation (#CNS-064371). We also thank the anonymous reviewers for their insightful comments and suggestions.


  1. Aitchison J, Brown J (1969) The lognormal distribution. Cambridge University Press, New YorkGoogle Scholar
  2. Akaike H (1982) Prediction and entropy. Tech. Rep. 2397, NTISGoogle Scholar
  3. Bishop P, Bloomfield R (2003) Using a log-normal failure rate distribution for worst case prediction. In: Proc. of the intl. symposium on software reliability engineering—industry practices, Colorado, pp 237–245Google Scholar
  4. Boehm B (1981a) Software cost estimation with COCOMO II. Prentice Hall, Englewood CliffsGoogle Scholar
  5. Boehm B (1981b) Software engineering economics. Prentice Hall, Englewood CliffsMATHGoogle Scholar
  6. Chillarege R (1996) Handbook of software reliability engineering. In: Lyu MR (ed) Orthogonal defect classification, chap. McGraw-Hill, New YorkGoogle Scholar
  7. Crow EL, Shimizu K (1988) Lognormal distributions: theory and applications. Marcel Dekker, New YorkMATHGoogle Scholar
  8. Dalal SR, Mallows CL (1990) Some graphical aids for deciding when to stop testing software. IEEE Trans Softw Eng 8(2):169–175Google Scholar
  9. Gokhale S, Mullen R (2004) From test count to code coverage using the lognormal failure rate. In: Proc. of intl. symposium on software reliability engineering (ISSRE), St. Malo, pp 295–395Google Scholar
  10. Gokhale S, Mullen R (2006) Queuing models for field defect resolution process. In: Proc. of intl. symposium on software reliability engineering (ISSRE), RaleighGoogle Scholar
  11. Gokhale S, Mullen R (2008) Software defect repair times: a multiplicative model. In: Proc. of predictor models in software engineering (PROMISE 08), LeipzingGoogle Scholar
  12. Gokhale S, Lyu MR, Trivedi KS (2004) Analysis of software fault removal policies using a non homogeneous continuous time Markov chain. Softw Qual J 12:211–230CrossRefGoogle Scholar
  13. Gokhale S, Lyu M, Trivedi KS (2006) Incorporating fault debugging activities into software reliability models. IEEE Trans Reliab 55:281–292CrossRefGoogle Scholar
  14. Hewett R, Kijsanayothin P (2008) On modeling software defect repair time. Empir Softw Eng 14(2):165–186CrossRefGoogle Scholar
  15. Johnson NL, Kotz S, Balakrishnan N (1994) Continuous univariate distributions. Wiley, New YorkMATHGoogle Scholar
  16. Keyes J (2003) Software engineering handbook. CRC, Boca RatonGoogle Scholar
  17. Kremer W (1983) Birth and death bug counting. IEEE Trans Reliab R-32(1):37–47MATHCrossRefGoogle Scholar
  18. Levendel Y (1990) Reliability analysis of large software systems: defect data modeling. IEEE Trans Softw Eng 16(2):141–152CrossRefGoogle Scholar
  19. Loeve M (1963) Probability theory. Van Nonstrand, New YorkMATHGoogle Scholar
  20. Miller DR (1985) Exponential order statistic models of software reliability growth. Tech. Rep. NASA Contractor Report 3909, NTISGoogle Scholar
  21. Mullen RE (1998a) The lognormal distribution of software failure rates: origin and evidence. In: Proc. of intl. symposium on software reliability engineering, pp 124–133Google Scholar
  22. Mullen RE (1998b) The lognormal distribution of software failure rates: applications to software reliability growth modeling. In: Proc. of intl. symposium on software reliability engineering, pp 134–142Google Scholar
  23. Mullen RE (2006) Characterizing software defect repair times. In: Proc. of the intl. symposium on software reliability engineering—industry practices, RaleighGoogle Scholar
  24. Mullen R, Gokhale S (2005) Software defect rediscoveries: a discrete lognormal model. In: Proc. of intl. symposium on software reliability engineering (ISSRE), pp 203–212Google Scholar
  25. NIST/SEMATECH (2006) e-handbook of statistical methods. http://www.itl.nist.gov/div898/handbook/apr163
  26. Patel JK, Read BC (1996) Handbook of the normal distribution. Marcel Dekker, New YorkMATHGoogle Scholar
  27. Petrov VV (1975) Sums of independent random variables. Springer, New YorkGoogle Scholar
  28. Sakamoto Y, Ishiguro M, Kitagawa G (1986) Akaike information criterion statistics. D. Reidel, DordrechtMATHGoogle Scholar
  29. Schneidewind N (2002) An integrated failure detection and fault correction model. In: Proc. of the intl. conference on software maintenance, pp 238–241Google Scholar
  30. Schroeder B, Gibson GA (2006) A large-scale study of failures in high-performance computing systems. In: Proc. of the intl. conference on dependable systems and networks, PhiladelphiaGoogle Scholar
  31. Stutzke MA, Smidts CS (2001) A stochastic model of fault introduction and removal during software development. IEEE Trans Reliab 50(2):184–193CrossRefGoogle Scholar
  32. Weiss C, Premraj R, Zimmerman T, Zeller A (2007) How long will it take to fix this bug? In: Proc. of fourth intl. workshop on mining software repositoriesGoogle Scholar

Copyright information

© Springer Science+Business Media, LLC 2009

Authors and Affiliations

  1. 1.Dept. of Computer Science and EngineeringUniv. of ConnecticutStorrsUSA
  2. 2.Corporate Quality Metrics, Cisco SystemsBoxboroughUSA

Personalised recommendations