A Requirements-Driven Approach for Building Architecture Projects of Self-Adaptive Systems

The proposal uses Palladio, e-Motions, Maude and SYBL in such a way that it enables expressiveness and flexibility in the specification, modeling, behavioral definition and performance analysis of self-adaptive models using a procedural approach. Such procedural organization has helped us to understand the formulation of each of the steps, which makes the approach flexible and consistently reproducible. We specify the elasticity requirements of the systems, allowing their adjustment. The verification of the specification and the adaptation in different components in the system allowed us to verify that previously specified strategies may be misleading and our approach proposes that the readjustment and analysis be carried out at design-time. With the development of the communication channel behavior rules, it is possible to consider the latency variation and adaptations considering the communication link. The procedure and the practical experience have allowed us to validate our hypothesis. The initial specifications, model and adaptations were tested at simulation time, and the results obtained through their use during a simulation were submitted for re-simulation and reanalysis, which led us to a more precise diagnosis of the system and, eventually, to a readjustment towards better results.A well-structured software architectural project allows us to analyze and describe the properties of a complex system towards a general and complete overview, thus enabling the alternative decision-making in the face of possible problems. However, the management of the changes in a Self-Adaptive System is a challenge that, in addition to complexity, has dynamism as its main feature. The possible modifications of its initial configurations in response to changes in the context must be considered, which makes the construction of these systems and the decision-making at design time difficult tasks, since the initial model can take different forms and configurations during the execution process. In this thesis, we identified that the use of a flexible and procedural approach can incorporate changes in the model that are necessary for a self-adaptive system, being useful to measure and model the unpredictability of these systems, since the composition and/or integration of different services require a detailed analysis of the adaptation choices. This approach differs from the others because it proposes the creation of models that consider and understand the changing nature of self-adaptive systems, from the possibility of testing them considering the constant change of quality parameters, requirements and context after adaptations occur, understanding that they can affect the system as a whole, even if applied to a specific component. This set of design-time decisions and modifications, such as specifying requirements that consider constraints, monitoring and adapting strategies, as well as modeling systems and their behavior, and system analysis, can be used as a starting point for the construction of systems with better quality of service results, since the behavior of the application is understood even when adaptations occur, still in the design phase