Dynamic distributed database over cloud environment

Abstract

Distributed Database System (DDS) typically consist of a number of interrelated databases (fragments) located at different geographic sites. These sites can communicate through a network and is managed by a distributed database management system (DDBMS). The most business organizations need more fixed servers to store their large databases that consist of very large amounts of data, which used by applications at different physical locations. These organizations will charge a lot to design the distributed database infrastructure of their system, especially in the beginning of the work. Cloud computing allows these organizations to tap into a virtual computing and storage resources over the internet and also benefit from utility like reliability, costs, scalability, as well as pay only for what they use. However, many emerging applications of distributed database systems generate very dynamic workloads with frequent changes in access patterns from different sites. Consequently, in this realistic dynamic environment, where the access probabilities of nodes to fragments and its replicas change over time, the optimum data re-allocation and replication of those fragments is the only way to increase the performance, efficiency, reliability and availability of the distributed database. The thesis first address a cluster based distributed and parallel database design over a cloud environment. The proposed architecture and its components are designed for parallel processing the client queries and allow users to access the distributed database from anywhere. It also allows vertical and horizontal fragmentation, allocation and replication decisions to be taken statically at the initial stage of designing the distributed database, without the need of empirical data about query executions. Moreover, it clusters the distributed database sites into disjoints clusters. Then, a dynamic re-allocation and replication algorithm called optimal fragment reallocation and replication (OFRAR) algorithm was proposed. Which allow migration and/or replication decisions to be taken by each cluster independently of other clusters. This makes it possible to use this algorithm without communication overhead or even using the proposed algorithm on all sites in the system. Finally,The thesis addresses two types of fragmentation. The first type of the fragmentation is the vertical fragmentation. In this thesis a Full Vertical Fragmentation, Allocation and Replication (FVFAR) scheme over the cloud environment was presented. The proposed scheme addresses the limitation of the previous vertical fragmentation solutions. It also provides vertical fragmentation, allocation and replication as a service over the cloud. The second type of fragmentation addressed in this thesis is the horizontal fragmentation. In this thesis an enhanced horizontal fragmentation, allocation and replication algorithm was proposed, which takes the horizontal fragmentation ,allocation and replication decisions at the initial stage of designing the distributed database.