Hybrid Optimization for Software Component Configuration
Abstract
Modern software distributions like e.g. the Eclipse platform or a Linux distribution are composed from a multitude of components
(packages).
Each of these components requires or disallows the installation of some other components. Computing admissible component configurations
that are optimal according to some criterion (as up-to-date as possible, minimal change of existing installation, ...) is
a challenging task - indeed, this problem has been shown to be NP-hard. The scope of this thesis is to look at how techniques
from techniques from constraint programming
(CP) can be used to improve existing models in integer programming (IP)
- the resulting search procedure will be a hybrid of CP and IP. For general pointers to the topic of component configuration
see e.g. the LoCoCo workshop series http://lococo.irill.org/.
