This study investigated the use of data streaming systems, focusing on continuous event ingestion and the influence of system configuration on performance. A reference model was presented, comprising three main stages - ingestion, processing, and storage - mediated by a message broker component. Based on this model, Apache Kafka was selected as the practical tool for implementation and experimental evaluation. Controlled tests were conducted by varying message sizes and producer parameters to measure system latency and throughput under burst-load scenarios. The results showed that tuning batch size and buffer memory significantly impacts system stability and performance. The work combines theoretical foundations with practical experimentation, documenting the effects of different configurations and providing guidance for designing pipelines in real-world applications.