A data management framework for secure and dependable data grid

Data grid enables the sharing and coordinated use of data from various sources and provides various services to fit the needs of high-performance distributed and data-intensive computing. In many data grid applications, security, availability, and access efficiency are important issues that need to be addressed. Also, some data grid applications involve mobile clients where mobile users access the information from hand held devices. Thus, how to minimize the communication cost for the mobile clients accessing the data in data grids should also be investigated.; The goal of this dissertation research is to develop a data management framework to facilitate secure, highly available, and highly efficient accesses in data grids, considering both mobile and wired clients. Replication is used as the basic approach to achieve high availability and efficient accesses. Data are replicated on storage servers that are close to the clients so that the communication cost is minimized. In the case of mobile environment, data are also replicated on the mobile clients themselves to improve the access efficiency. To achieve security protection in data grids, secret sharing and erasure coding (with encryption) schemes are used to partition the data into data shares and distribute them on different storage servers. Thus, the security of the data can be assured even if some storage servers in the data grids are malicious or compromised. To improve the access performance of the partitioned data, data shares are further replicated and placed close to the clients. In both cases, how to place the data objects is the key issue in performance improvement and availability and security enhancement. Thus, this dissertation focuses on the data placement issues for correlated data due to transaction accesses and due to data partitioning.; In this dissertation research, several data placement algorithms for correlated data have been developed. Some of these algorithms determine the data placement solutions that can optimize the access efficiency. Some algorithms further consider the conflicting security, availability, and access efficiency objectives and aiming at satisfying security, availability, and efficiency constraints while optimizing these objectives. Assessment approaches have also been developed so that the security, availability, and performance objectives can be quantitatively measured to guide the placement decisions. These data placement algorithms and assessment approaches servers as the core of the data management framework. Future research efforts will focus on the development of innovative techniques for the remaining parts of the framework and building an ecological data grid using these techniques.; The research results from this dissertation provide the technologies for developing secure, highly available, and highly efficient data grid applications. This is especially beneficial for critical data grid applications that have stringent dependability requirements.