Objektové úložiště v Linuxu

Abstract

The thesis examines the architecture and deployment of open-source object storage systems, which offer the necessary scalability for managing large volumes of unstructured data compared to traditional systems. The goal is to theoretically describe their principles, select and compare two representative systems based on research, install them, and demonstrate their advanced features in~practice. A~MinIO server and a Ceph cluster are implemented in~a~virtualized Linux environment, verifying Ceph's autonomous self-healing capability.~Three practical use cases are then executed using the~S3~protocol. These demonstrate the successful deployment of object storage for the Nextcloud application, the~parallelized import of image data with transformation into intelligent objects enriched with clinical metadata, and the integration of an independent evaluation layer with a~classification model. This model analyzes streamed X-ray images directly in main memory and writes the results back into the object metadata. The findings confirm that object storage architecture overcomes the~limitations of~traditional systems, providing a highly reliable infrastructure for secure data archiving, cloud-native application development, and~machine learning model integration.