Doctoral Thesis Defense: Azadeh Jahanbanifar

Speaker: Azadeh Jahanbanifar

Supervisors: Drs. F. Khendek, M. Toeroe

Examining Committee:
Drs. J. Bentahar, D. Goswami, J. Rilling, D. Petriu,
R. Dssouli (Chair)

Title: A Model-based Framework for System Configuration Management

Date: Thursday, March 31, 2016

Time: 13:00

Place: EV 1.162

ABSTRACT

A system can be viewed from different perspectives, each focusing on a specific aspect such as availability, performance, security. Configurations reflect the manageable resources of the system, their attributes and organization which are necessary for the management of the system for each aspect. Thus, for management purposes a system is generally described through various partial configurations (also known as configuration fragments). To form a consistent system configuration, these independently developed configuration fragments need to be integrated together. The integration of configuration fragments is a challenging task. This is mainly due to overlapping entities (different logical representations of the same system resource) in the configuration fragments and/or complex relationships among the entities of the different configuration fragments. At runtime the system may be reconfigured to meet certain/new requirements or in response to performance degradations. These changes may lead to inconsistency as some changes may violate the constraints between entities. Maintaining the consistency and adjusting the system configuration at runtime is another challenging task.

In our research, we propose to handle these two important issues in an integrated manner. We define a model-based framework for configuration management. Using model weaving and model transformation techniques, we propose a solution for the integration of configuration fragments targeting specific system properties. To handle runtime changes, we propose a configuration validation and adjustment solution to check and preserve the consistency of the system configuration. The runtime reconfigurations are checked against a reduced set of consistency rules and the reconfigurations are applied only if they are safe. For handling the changes that violate the consistency rules, we propose an adjustment technique to automatically resolve (if possible) the inconsistencies. This is achieved by propagating the changes in the configuration according to the system constraints following the possible impacts of the configuration entities on each other. Some heuristics are used to control the propagation and reduce the complementary changes and adjustment time.