Efficient replication of large volumes of data and maintaining data consistency by using P2P techniques in Desktop Grid

Desktop Grid is increasing in popularity because of relatively very low cost and good performance in institutions. Data-intensive applications require data management in scientific experiments conducted by researchers and scientists in Desktop Grid-based Distributed Computing Infrastructure (DCI). Some of these data-intensive applications deal with large volumes of data. Several solutions for data-intensive applications have been proposed for Desktop Grid (DG) but they are not efficient in handling large volumes of data. Data management in this environment deals with data access and integration, maintaining basic properties of databases, architecture for querying data, etc. Data in data-intensive applications has to be replicated in multiple nodes for improving data availability and reducing response time. Peer-to-Peer (P2P) is a well established technique for handling large volumes of data and is widely used on the internet. Its environment is similar to the environment of DG. The performance of existing P2P-based solution dealing with generic architecture for replicating large volumes of data is not efficient in DG-based DCI. Therefore, there is a need for a generic architecture for replicating large volumes of data efficiently by using P2P in BOINC based Desktop Grid.
Present solutions for data-intensive applications mainly deal with read only data. New type of applications are emerging which deal large volumes of data and Read/Write of data. In emerging scientific experiments, some nodes of DG generate new snapshot of scientific data after regular intervals. This new snapshot of data is generated by updating some of the values of existing data fields. This updated data has to be synchronised in all DG nodes for maintaining data consistency. The performance of data management in DG can be improved by addressing efficient data replication and consistency. Therefore, there is need for algorithms which deal with data Read/Write consistency along with replication for large volumes of data in BOINC based Desktop Grid.
The research is to identify efficient solutions for data replication in handling large volumes of data and maintaining Read/Write data consistency using Peer-to-Peer techniques in BOINC based Desktop Grid. This thesis presents the solutions that have been carried out to complete the research.