Predicting quality attributes of software product lines using software and network measures and sampling

  • Authors:

  • Sergiy S. Kolesnikov;Sven Apel;Norbert Siegmund;Stefan Sobernig;Christian Kästner;Semah Senkaya

  • Affiliations:

  • University of Passau, Germany;University of Passau, Germany;University of Magdeburg, Germany;Vienna University of Economics and Business, Austria;Carnegie Mellon University;University of Munich, Germany

  • Venue:
  • Proceedings of the Seventh International Workshop on Variability Modelling of Software-intensive Systems
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

Software product-line engineering aims at developing families of related products that share common assets to provide customers with tailor-made products. Customers are often interested not only in particular functionalities (i.e., features), but also in non-functional quality attributes, such as performance, reliability, and footprint. Measuring quality attributes of all products of a product line usually does not scale. In this research-in-progress report, we propose a systematic approach aiming at efficient and scalable prediction of quality attributes of products. To this end, we establish predictors for certain categories of quality attributes (e.g., a predictor for high memory consumption) based on software and network measures, and receiver operating characteristic analysis. We use these predictors to guide a sampling process that takes the assets of a product line as input and determines the products that fall into the category denoted by the given predictor (e.g., products with high memory consumption). We propose to use predictors to make the process of finding "acceptable" products more efficient. We discuss and compare several strategies to incorporate predictors in the sampling process.