Titulo Estágio
Prediction and Rejuvenation of Software Aging in Containerized Environments
Áreas de especialidade
Engenharia de Software
Local do Estágio
SSE
Enquadramento
Long-running services deployed via container technologies like Docker and Podman are subject to performance degradation over time, a phenomenon known as software aging. This includes memory leaks, increased CPU usage, and fragmentation, which can compromise service availability and increase operational costs. Despite increasing adoption of containers in cloud and edge computing, aging detection and mitigation techniques in such platforms remain underexplored.
Objetivo
This project aims to develop an intelligent approach for predicting software aging effects in containerized environments and performing selective rejuvenation actions to avoid system degradation. The solution will leverage machine learning techniques, to forecast resource exhaustion based on monitoring data. When degradation is anticipated, automatic rejuvenation actions such as restarting affected containers will be triggered.
Plano de Trabalhos - Semestre 1
T1. [15/09/2025 to 15/10/2025] Study the problem domain. Review of the state of the art in software aging, rejuvenation, and ML-based prediction techniques.
T2. [15/10/2025 to 15/11/2025] Requirements and technology selection. Define the metrics, monitoring stack, data labeling strategy, and architecture of the predictive system.
T3. [15/11/2025 to 31/12/2025] Approach specification and tool preparation.
a. Define the prediction model (e.g., MLP inputs, architecture, training targets).
b. Extend the aging benchmark to generate training data, simulate aging, and apply rejuvenation policies.
T4. [01/01/2026 to 21/01/2026] Write the Dissertation Plan.
Plano de Trabalhos - Semestre 2
T5. [01/02/2026 to 30/03/2026] Implementation and initial experiments.
a. Train and validate the MLP model for aging prediction based on benchmark data.
b. Integrate the predictor with a decision engine for automatic rejuvenation.
c. Conduct benchmark runs with and without rejuvenation.
T6. [01/04/2026 to 30/04/2026] Evaluation.
a. Assess the effectiveness of the predictive rejuvenation approach versus baseline.
b. Measure system stability, resource efficiency, and performance under long workloads.
T7. [15/04/2026 to 31/05/2026] Write a paper or technical report.
T8. [15/05/2026 to 01/07/2026] Write the thesis.
Condições
The selected candidate will be integrated into the research team of the Software and Systems Engineering Group at CISUC. There is a possibility of a scholarship. The work will be developed in CISUC laboratories, with full access to computing resources and workspace.
Orientador
João Ferreira da Silva Júnior
jfjunior@dei.uc.pt 📩