USAGE OF APACHE KAFKA FOR LOW-LATENCY IMAGE PROCESSING
Nazar Karpiuk, Halyna Klym, T. Tkachuk · Electronics and Information Technologies · 2024
This study addresses the limitations of conventional centralized systems in handling the surge in real-time image processing demands. We propose a distributed architecture employing Apache Kafka to achieve near real-time image analysis. Our approach implements a decoupled workflow for image acquisition, processing, and spreading, facilitating parallel execution across a processing node cluster. Kafka acts as the core communication and data flow infrastructure, ensuring scalability, fault tolerance, and high throughput. Evaluations demonstrate substantial performance gains compared to a centralized system, validating the feasibility, advantages, and limitations of Kafka for distributed image processing. We systematically analyzed the impact of topic partitioning, consumer group configuration, and processing workload on performance. This work presents a robust solution for near real-time image processing tasks, promoting the development of efficient and scalable image analysis applications. Key words: Apache Kafka, image processing, distributed environment, parallel processing.