An Approach for Robust Distributed Data Retrieval in Data Intensive Grid Environments

摘要

Data Intensive Grid is meant for efficient data sharing in large scale distributed data storage environment, which is a promising way to improve data and software viability. Among numerous aspects which needs consideration, data availability is one of the most vital and difficult, since it faces lots of challenges such as abnormal failures of nodes, dynamic nodes participation and unstable network connection. Adding data replicas in the Data Grid environment can improve data availability and response time and thus ensure robust data retrieval. Yet how to replicate data in order to meet performance requirements in data Grid still remains unsolved. In this paper we propose a solution that stripes large files into data blocks and distributes over data nodes for storage and design a mechanism for automatic data replication of nodes aiming at ensuring data availability. Preliminary simulation experiments show that our approach is effective in improving data availability with limited increase of system costs.