Toward a Software Testing and Reliability Early Warning Metric Suite

  • Authors:

  • Nachiappan Nagappan

  • Affiliations:

  • North Carolina State University

  • Venue:
  • Proceedings of the 26th International Conference on Software Engineering
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

The field reliability is measured too late for affordablyguiding corrective action to improve the quality of thesoftware. Software developers can benefit from an earlywarning of their reliability while they can still affordablyreact. This early warning can be built from a collection ofinternal metrics. An internal metric, such as the numberof lines of code, is a measure derived from the productitself [15]. An external measure is a measure of a productderived from assessment of the behavior of the system[15]. For example, the number of defects found in test isan external measure. The ISO/IEC standard [15] statesthat "[i]nternal metrics are of little value unless there isevidence that they are related to external quality."Internal metrics can be collected in-process and moreeasily than external metrics. Additionally, internalmetrics have been shown to be useful as early indicatorsof externally-visible product quality [1]. For these earlyindicators to be meaningful, they must be related (in astatistically significant and stable way) to the fieldquality/reliability of the product. The validation of suchmetrics requires the convincing demonstration that (1) themetric measures what it purports to measure and (2) themetric is associated with an important external metric,such as field reliability, maintainability, or fault-proneness[12].Software metrics have been used as indicators ofsoftware quality [1, 19-21, 23] and fault proneness [8-10,24]. There is a growing body of empirical results thatsupports the theoretical validity of the use of higher-orderearly metrics, such as OO metrics [1] defined byChidamber-Kemerer (CK) [6] and the MOOD [5] OOmetric suites as predictors of field quality. However,general validity of these metrics (which are oftenunrelated to the actual operational profile of the product)is still open to criticism [7].