The application of product measures in directing software maintenance activity

Abstract

This paper is concerned with enhancing aspects of project‐management techniques in relation to maintenance in commercial object‐oriented software. Maintenance forms a significant proportion of the lifecycle of a product and consequently its costs. Techniques such as lean production attempt to reduce cost by focusing team effort and to pre‐empt issues by the use of early detection and resolution strategies. This study involves a commercial application written in C++ and its evolution over a one‐year period. Software product measures are used in conjunction with Pareto analysis to determine effective predictive class targeting in a release cycle. We assess the capability of the measures and consider how the results can be used by the project team. The study extends the work of others in this field in three ways. First, other studies have analysed procedural code, whereas we use an object‐oriented code base. Second, most studies assess the predictive capability of measures against the number of software issues; we examine these factors, but uniquely test for change density using density modified measures and we also test for revision volume. Finally, we examine the implications of these results and offer a cohesive framework of recommendations placed within the lean management philosophy. Copyright © 2007 John Wiley & Sons, Ltd.

Fault detection and prediction in an open-source software project

“…It is a principle that seems to holds in many different contexts and was orignally used to explain income and land ownership distributions. It has also been shown to hold in the field of software engineering where a number of empirical studies have shown that a large proportion (80%) of faults and changes are centered in a small proportion (20%) of classes [40,55]. Koru and Liu [40] showed that 80% of the changes to KOffice and Mozilla were centered in 20% of the classes.…”

“…Koru and Liu [40] showed that 80% of the changes to KOffice and Mozilla were centered in 20% of the classes. Ware et al [55] examined the evolution of a commercial application over one year and found that Pareto analysis is useful for predicting change-prone classes. Given the necessity outlined in the introduction to focus testing effort on the parts of the software that are most likely to be faulty, Pareto's Law can help to focus the testing effort on these parts of the software.…”

A model of open source software maintenance activities

“…Second, we derive the release time of the newer version based on our model. A graphic illustration of the fitted values is given in Fig 1. It is observed that LSE curve is very close to the actual curve in the intervals [0, 4] and [13,27] while it is far from the actual curve in the interval [5,10]. MLE curve generally represents the pattern of the actual curve.…”

“…Issues on software maintenance in traditional closed source software systems, such as scheduling and strategy [3,10], methods [4,11], policy modeling [5,6], etc. have been systematically studied by many authors.…”

Optimising agile development practices for the maintenance operation: nine heuristics

“…Much research in software maintenance has been focused on traditional software engineering techniques such as modelling techniques and UML, (Arisholm et al 2006;Dzidek et al 2008), estimation, (Buchmann et al 2011;De Lucia et al 2003;Nguyen et al 2011), risk management, (Charette et al 1997;Sherer 1997), statistical process control (De Lucia et al 2003;Popovic et al 2001;Ware et al 2007;Zanker and Gordea 2006), quality, (Ghods andNelson 1998), metrics, (Hall andLineham 1997;Popovic et al 2001;Ware et al 2007), post mortem, (De Sousa et al 2004) and testing, (Sukumaran and Sreenivas 2005). The major thrust of this research -in line with standard textbooks such as Pigoski (1996) and Grubb & Takang (2003) -is directed towards increasing discipline (Boehm and Turner 2003) in the maintenance operation, rather than agility.…”