Daniel Lehner,
"Model-based Detection of Runtime Inconsistencies"
, 9-2020
Original Titel:
Model-based Detection of Runtime Inconsistencies
Sprache des Titels:
Englisch
Original Kurzfassung:
With an increasing dependency on software in more and more aspects of our lives, theinformation whether these software systems work as expected at any point in time becomescrucial. However, limitations of existing approaches lead to a loss of information aboutconsistency between expected and actual behavior, especially after system deployment.(1) Full validation of software artifacts is often not possible any longer in modern softwaresystems. Increasing complexity leads to an exploding number of execution possibilities,which can be infeasible to compute and check. (2) Software testing allows to overcomethis challenge by allowing to check particular parts of a system, but with limitations whenit comes to consistency after system deployment. (3) Runtime monitoring systems provideinformation on system errors after deployment, but the high manual effort required toset up such a system limits its application in practice. In this thesis, it is examined towhich extent reusing existing model information can provide a solution to the challengesfaced with the approaches mentioned above. Therefore, the Inconsistency DetectionBased on Models (IDBoM) framework is implemented as a showcase how existing modelinformation can be used for inconsistency checking. The IDBoM framework uses UMLActivity Diagrams and UML Class Diagrams as a representation of the expected behaviorof a system, and UML Sequence Diagrams as representation of its actual behavior afterdeployment. After the inconsistency checking process is triggered by a running system, aset of consistency rules is executed on these diagrams to retrieve a checking result. Thischecking result indicates whether the actual behavior of the checked system is consistentwith its expected behavior. In order to facilitate the consistency between expectedsystem behavior and existing models, a reusable solution for programatically interactingwith models with the focus on usability is created as part of the IDBoM framework.The intention of this solution is to contribute to the implementation of use cases forreusing existing model information, to increase benefits of keeping models up to date.An evaluation of the implemented artifact shows that full automation of a model-basedinconsistency checking process is possible, from process initiation to processing of thechecking result. The covered inconsistency cases are comparable to software testing.Execution times of the implemented inconsistency checking process scale linearly for alltested parameters. Finally, the usability of model interactions is improved compared toexisting solutions on all tested parameters. These results indicate that a model-basedinconsistency checking process can be used to provide information about the correctbehavior of a system after deployment
Forschungs-