Intelligent optimization of distributed systems: performance analysis with Lithops and machine learning

Identifier:  TFM:2107

Handle: https://hdl.handle.net/20.500.11797/TFM2107

Authors:  Benabdelkrim Zakan, Usama

Abstract:
This thesis presents a solution for optimizing and monitoring distributed systems using Lithops in serverless environments. A lightweight profiler integrated with Prometheus was developed to collect real-time metrics and manage resources efficiently. The profiler tracks CPU, memory, disk, and network usage, ensuring scalability and resilience across multicloud platforms. Additionally, a machine learning model predicts the optimal task parallelization to minimize execution time. The proposed solution was validated through extensive testing in various environments, demonstrating its effectiveness in improving serverless computing performance by offering a robust tool for enhancing operational efficiency and resource optimization.