| During the past few decades there has been a significant growth in data-driven applications that heavily rely on information derived from many data sources. The services of these applications are often powered by a backend data repository, frequently a relational database with information that is from many data sources. The constantly increasing number of users for these applications results in a large volume of queries that need to be supported.;This dissertation studies quality issues in client-server database applications. Quality issues can be related, for example, to minimization of resource contention while guaranteeing high-quality results. They can also be related to data-quality modeling, high query throughput in the system, and more.;This dissertation focuses on problems that arise in the maintenance of the local data repository, as well as efficient query processing within the repository. Specifically, a viable repository must be efficient, incorporating resolving mechanisms for resource contention, and optimizing algorithms in its core design. Because the data in the majority of sources is dynamic, repositories require efficient mechanisms for retrieving updates. The retrieving update processes are permanent features in the repository life-cycle, running in parallel to the user query-execution processes. An increasing number of user queries and rapidly changing data generate contention between the two processes on the network and computational resources. Moreover, efficient query processing is a crucial requirement for huge amounts of collected data and a large volume of queries, as in client-server applications. Users issue queries to the client which in turn makes query execution plans. The client should support many user requests, and in order to guarantee a certain level of query throughput it must efficiently process the requests. We also propose several algorithmic solutions that provide best-effort query processing in the presence of limited resources. |
PDF Full Text Request