MAFC: Multimedia Application Flow Controller for Big Data Systems

A lot of research has been conducted on different areas of Big Data technology in the multimedia domain, such as multimedia Big Data capture, storage, indexing and retrieval. Nevertheless, ensuring real-time processing for multimedia Big Data is still a big challenge. Indeed, when the streaming speed is higher than the processing time, data can be lost or the processing no longer respects real-time constraints even though efficient streaming technology is used. In this context, we propose MAFC, a Multimedia Application Flow Controller that filters incoming multimedia data so that only significant data is sent for processing. Data extraction is executed following application-based policies. MAFC is a building block that can be embedded in the management layer of Big Data systems. The main contribution of our work is a Big Data architecture that couples multimedia data filtering with the Big Data streaming technology in order to reduce the stress on the processing layer. As a use-case, the integration of MAFC in a Big Data-based surveillance system for video frame extraction shows a significant performance improvement compared to the initial system.