An Investigation of Integration Debt in Continuous Integration - A design science study
Typ
Examensarbete för masterexamen
Program
Publicerad
2019
Författare
Rasku, Kevin
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
Continuous integration is a commonly used method in software development. It is
used to improve the development process by prescribing frequent code integration
and testing and thereby creating a mechanism for early feedback. To optimize
the use of continuous integration in large scale development in regards to code
integration, this study introduced a new term called integration debt. The concept
is borrowed from the notion of technical debt, applied to continuous integration
and the factors that are involved in this integration debt (which happens due to
unintegrated code) are investigated. In this study, a set of integration debt factors is
defined and the relationships between the factors are mapped and analyzed. This is
done through design science research methodology where the artifact is the mapping
of integration debt factors and the set of factors itself. Using various data gathering
methods, a set of factors that are involved in integration debt are identified and
validated. These factors were lines of code in commits, the number of modified
files in commits, code churn, frequency of commits, code dependency, development
overlap and lastly the factors related to the organization’s CI system.
Additionally, this study investigates, and evaluates, a potential threshold value for
one of the factors where the threshold is for when that factor reaches a significant
risk to integration. Defining a general and accurate threshold proved to be difficult
which suggests that integration debt factor thresholds require significant effort to
define. This could perhaps be investigated further in future research. Consequently,
the findings of this research study build a foundation for integration debt and how
it could be mapped and measured. This can aid companies that work with large
scale development, to better manage the integration debt level across their project
by taking appropriate actions to avoid large accumulated debt levels. This potentially
lowers the costs for companies and reduces the extra effort required to resolve
integration problems during development.
Beskrivning
Ämne/nyckelord
Continuous integration , Integration debt , Repository mining , Software metrics , Code integration , Merge conflict , Automated test and build