Improving Transaction Acceptance of Incoherent Updates Using Dynamic Merging In a Relational Database
Despite its tenure, mobile computing continues to move to the forefront of technology and business. This ever expansive field holds no shortages of opportunity for either party. Its benefits and demand are abundant but it is not without its challenges. Maintaining both data consistency and availability is one of the most challenging prospects for mobile computing. These difficulties are exacerbated by the unique ability of mobile platforms to disconnect for extended periods of time while continuing to function normally. Data collected and modified while in such a state poses considerable risk of abandon as there exists no static algorithm to determine that it is consistent when integrated back to the server. This thesis proposes a mechanism to improve transaction acceptance without sacrificing consistency of the related data on both the client and server. Particular consideration is placed towards honoring data which a client may produce or modify while in a disconnected state. The underlying framework leverages merging strategies to resolve conflicts in data using a custom tiered dynamic merge granularity. The merge process is aided by a custom lock promotion scheme applied in the application layer at the server. The improved incoherence resolution process is then examined for impacts to the fate of such transactions and related bandwidth utilization.
Table of Contents
Introduction -- Related work -- Approach -- Implementation -- Evaluation -- Conclusion -- Appendix A. Client API documentation -- Appendix B. Client code -- Appendix C. Server code -- Appendix D. Property performance data