Scalable continuous query processing in location -aware database servers
The wide spread use of cellular phones, handheld devices, and GPS-like technology enables location-aware environments where virtually all objects are aware of their locations. Location-aware environments and location-aware services are characterized by the large number of moving objects and large number of continuously moving queries (also known as spatio-temporal queries). Such environments call for new query processing techniques that deal with the continuous movement and frequent updates of both spatio-temporal objects and spatio-temporal queries. This dissertation, presents novel paradigms and algorithms for efficient processing and scalable execution of continuous spatio-temporal queries in location-aware database servers. We introduce a disk-based framework that exploits shared execution and incremental evaluation paradigms. With shared execution, the problem of evaluating a set of concurrent continuous queries is abstracted to a spatial join between the set of moving objects and the set of moving queries. With the incremental evaluation, rather than performing a repetitive evaluation of continuous queries, we produce only the updates of the recently reported answer. For streaming environments, we introduce a generic class of spatio-temporal operators that can be tuned with a set of parameters and methods to act as various continuous spatio-temporal queries (e.g., range queries and k-nearest-neighbor queries). The spatio-temporal operators can be combined with other traditional operators (e.g., join, distinct, and aggregate) to support a wide variety of continuous spatiotemporal queries. To support scalability in steaming environments, we introduce a salable operator that shares memory resources among all outstanding continuous queries. To cope with intervals of high arrival rates of data objects and/or continuous queries, the proposed scalable operator utilizes a self-tuning approach based on load-shedding where some of the stored objects are dropped from memory. The experimental evaluation of our disk-based approach compares with recent scalable approaches and shows the superior performance of our techniques. Also, we experimentally evaluate our spatio-temporal operators based on a real implementation inside an open-source data stream management system. The experimental results show that by delving inside the database engine and providing pipelined operators for continuous spatio-temporal queries, we can achieve performance orders of magnitude better than other application level algorithms.
Aref, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our