Using models at run-time to measure quality of SAS in the large-scale software systems

Abstract

[Context/ Motivation] The adaptation process in self-adaptive software systems modifies the system based on new monitored conditions to make the system able to deserve SLA (QoS). Some Researchers assume the environment as a closed-world. So, adaptation actions are predicted at design-time, to get applied at run-time. But uncertainty causes prediction about the all different conditions be impossible. [Objective] In this study we aim to handle the uncertainty consequences in the open-world by the use of two quality models presented in the process level and the product level. [Method] To reduce the cost of adaptation at run-time, an integrated approach for modeling and verifying the requirements of SAS is used as well as a measurement method to measure the level of satisficing quality factors using Model Driven Engineering (MDE). [Results] When the measurement (i.e. the deviation from desired behavior) is taken into account in early phases of development, not only the adaptation cost degrades, but also the undesired consequences do not propagate to the next phases. [Conclusion] To sum up, in this paper requirements are modeled based on GORE models. Then an SAS Quality model is augmented in the goal model. This embedded model is verified by OMEGA2/ IFX profile. The verified model is transformed to alternative architectural models by some defined transformation rules which is written in ATLAS language, implemented in Eclipse Modeling Framework (EMF).