Update and transaction processing in peer data sharing systems

Title: Update and transaction processing in peer data sharing systems
Authors: Masud, Md. Mehedi
Date: 2009
Abstract: Our thesis investigates an update exchange mechanism in data sharing systems. In these systems, databases in peers are created independently and may have semantic inter-dependencies with regards to data. Therefore, each peer specifies pair-wise mappings with other peers for sharing and exchanging related data. The mappings express how data in one peer relate to data in another peer. In our thesis, we first define settings for sharing and exchanging related data in a data sharing system, and then we discuss an update exchange semantics considering these settings. The update exchange models how changes made to a peer's instance are applied to the peer's local database instance and then propagated to related peers for maintaining data consistency in data sharing systems. The propagation is subject to transformations of the updates wrt the schema and data vocabularies (i.e., constants and relation naines) of the related peers, and the results of the updates are consistent wrt the defined data sharing settings. Second, we propose an update translation strategy between a peer and its acquainted peers. The translation strategy first filters those updates from consideration that are ineligible for execution at the acquainted peers. During the translation, the locally expressed updates are transformed in terms of the schema of acquainted peers. Moreover, the translation ensures that insertions, deletions, and modifications of the tuples made by an update in a peer and by the translated version of the update in an acquainted peer are related through the mappings between both peers. The translation is performed by considering the mappings between peers and through a syntactic analysis of update operations. Third, we investigate an update conflict detection and resolution mechanism during the propagation of updates. Fourth, we investigate the processing of transactions in a data sharing system. We mainly focus on how to maintain a consistent execution view of concurrent transactions in peers without a global transaction coordinator. For this purpose, we investigate potential problems that arise when maintaining a consistent execution of concurrent transactions. In order to guarantee consistent execution, we introduce a correctness criteria and propose two approaches, namely the Merged Transactions method and the Ticket method. We assume that one single peer initiates the concurrent transactions. In this thesis, we also discuss the architecture and functionalities of a simulation tool, called PDST, that we developed for simulating large peer data sharing systems. We evaluate our proposed strategies using this tool.
URL: http://hdl.handle.net/10393/29943
CollectionTh├Ęses, 1910 - 2010 // Theses, 1910 - 2010
NR61388.PDF3.96 MBAdobe PDFOpen