الفهرس | Only 14 pages are availabe for public view |
Abstract Many recent applications in several domains such as sensor networks and financial applications generate continuous, rapid, and time varying datasets which are called data streams. Data streams require real time processing. In most database systems, the query optimizers select a single plan to process all streams tuples which is not efficient with the streams changeable nature. Also there is a little research effort has been made towards the multiple data streams queries’ simultaneous execution. In addition applying data streams’ multi-directional optimization over an optimized and elastic environment has not been much considered. Thus in this thesis we proposed combined frameworks and different optimization algorithms to solve these problems. First, we proposed the optimized query mesh for data stream (OQMDS) framework. In which data streams are processed over multiple query plans. Each plan is used to process a cluster of data that have nearest properties. We proposed the Optimized Iterative Improvement Query Mesh (OII-QM) and Non-Search based Query Mesh (NS-QM) algorithms, to efficiently generate the multiple plans. The proposed algorithms improves the optimization time by 70.3%, the execution time by 21.8%, the execution overheads by 80% and the memory usage by 96% over the II-QM algorithm. Then in this thesis the Continuous Query Optimization based on Multiple Plans framework for data streams over the cloud environment (CQOMP) was proposed. CQOMP provides an optimized streams processing over the cloud. The Optimized Multiple plans (OMP) and the Auto Scaling Cloud Query Mesh (AS-CQM) algorithms were proposed for streams processing over multiple query plans on cloud computing. The proposed OMP improves the performance in terms of the execution time by 83.5%, 47.7%, and the throughput by 69.7%, 40% over the operator-set-cloud methodology and the NS-QM algorithm. The elastic configurations of the proposed AS-CQM increases utilizing cloud processing resources by 33.8% and reduces the costs by 50% over the static configuration. Finally in this thesis the multiple queries optimization based on partitioning (MQOP) framework was proposed to efficiently execute multiple queries simultaneously on the cloud environment. The optimized global plan (OGP) and the optimized global plan based on partitioning (OGPP) algorithms were proposed for jointly executing multiple continuous queries over an optimized global plan to each cluster of data on the cloud. The proposed OGP improves the execution time by 80% and the throughput by 76.8% over the operator tree technique. The proposed OGPP algorithm improves the performance in terms of execution time by 61.1%, 72.6%, and the throughput by 55.5%, 66.5% over the compile time optimization method and the operator tree technique. |