Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
DISCENTE: LUIS GUILHERME SOUSA E SILVA
DATA: 28/05/2025
HORA: 09:00
LOCAL: Plataforma Google Meet
TÍTULO: Availability and Performance Evaluation of VANETs Based on Stochastic Petri Nets
PALAVRAS-CHAVES: Vehicular Ad Hoc Networks; Edge Computing; Microservices; Stochastic Petri Nets; Availability; Performance; Microservice Migration; Autoscaling; Reinstantiation
PÁGINAS: 80
GRANDE ÁREA: Ciências Exatas e da Terra
ÁREA: Ciência da Computação
SUBÁREA: Sistemas de Computação
ESPECIALIDADE: Software Básico
RESUMO:

The evolution of urban mobility requires computational solutions that ensure service continuity, adaptability and resilience in the face of operational variations and runtime failures. In this context, vehicular ad hoc networks (VANETs), integrated with edge computing and microservices architecture, play a central role in distributed systems for traffic monitoring and control. However, topological instability, resource limitations and high validation costs with real experiments make it essential to use analytical approaches for prior reliability and performance assessment. This dissertation proposes models based on Stochastic Petri Nets (SPNs) to represent and analyze the availability and performance of a VANET architecture with multiple RSUs connected to an edge server. The availability model incorporates physical, logical and communication failure modes, focusing on operational continuity ensured by microservice migration between RSUs. The modeling considers parameters such as Mean Time Between Failures (MTTF) and Mean Time to Repair (MTTR). The case studies show that, even in the face of simultaneous failures in different system components, activating the migration mechanism prevents complete degradation of availability and ensures the maintenance of operational paths, with faster recovery and lower risk of functional collapse. For performance evaluation, an SPN model was developed that represents the adaptation of processing capacity in RSUs through horizontal autoscaling and microservices reinstantiation. The model considers different traffic volumes, variations over time, queue size, and initial resource availability. The case studies reveal saturation patterns that occur when scaling limits are not well adjusted, and indicate that the combined use of autoscaling with reinstantiation reduces bottlenecks and improves the balance between allocated capacity and actual demand. The evaluation also showed that small variations in parameters such as the initial number of microservices or queue depth have a significant impact on response and discard metrics. Sensitivity analysis via Design of Experiments (DoE) identified the most influential factors, allowing guidance on the optimal configuration of operational policies.

MEMBROS DA BANCA:
Presidente - 1042211 - FRANCISCO AIRTON PEREIRA DA SILVA
Interno - 1732264 - JULIANA OLIVEIRA DE CARVALHO
Interno - 1732813 - RAYNER GOMES SOUSA

Cadastrada em: 25/04/2025