Simulation-based Optimization of System-of-Systems Software Architectures

• Autor(a) principal: Manzano, Wallace Alves Esteves
• Data de Publicação: 2023
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/55/55134/tde-15062023-163024/
• topic: Aprendizado de máquina; Arquitetura de software; Machine learning; Optimization; Otimização; Sistemas-de-Sistemas; Software architecture; Systems-of-Systems

Society is increasingly dependent on services provided by complex, integrated software systems so that an isolated system has not been successful in meeting these demands. In this scenario, Systems-of-Systems (SoS) have emerged as a result of the interoperability of independent systems both operationally and managerially to provide more complex solutions that no system would be able to provide in isolation. Due to the independence of the SoS constituent systems, these constituents can exit and enter the SoS at any time, resulting in a highly dynamic architecture. SoS is often linked to critical tasks, that is, which can pose threats to human integrity. Thus, the system must ensure that the software architecture has a high level of quality so that the system can offer a reliable and flawless service. However, due to the non-determinism generated by dynamic architecture, it is necessary to check the architecture at design time. Considering the high cost and threats of implementing critical systems without due verification of their architecture, this masters project proposes a simulation-based optimization framework using meta-heuristics to enable the optimization of quality attributes, including performance and availability, of SoS software architectures. This framework consists in two main elements, a process that defines how the simulation-based optimization is performed in the context of SoS software architecture, and an infrastructure that defines the software requirements to perform the process. This framework was evaluated using a case study of a smart building, and results shows that it is successful in carrying out with a good performance the task of performing simulation-based optimization of a SoS software architecture.