Exploiting Storage Redundancy and Parallelism for Efficient Retrieval of Multimedia Data

点击翻译按钮获取中文摘要

With the rapid advances in computing and communication technologies, there is increasing demand for and reliance on storing a large variety of data types in digital form. Once data is stored digitally, it becomes available for access and retrieval using a variety of parameters. This project addresses the problem of retrieval of multimedia data, which tends to be very large and hence requires a vast amount of storage. To capture the information content of data objects succinctly, it is represented as points/vectors in a multidimensional data space. Methods for efficient placement and retrieval mechanisms for multidimensional data, based on exploiting redundant and parallel storage, are developed in this project. The role of data partitioning/declustering and its interaction with indexing to support efficient retrieval of multidimensional data is explored. Efficient partitioning/declustering techniques for retrieving multimedia data and NASD (networked attached secure disk) and active disk architectures developed during this proposal will be integrated with the NSF sponsored Alexandria Digital Library project at University of California at Santa Barbara, and the results are also expected to have a wide range of applications in multimedia information systems.