| Over the last few years, database systems have expanded into new application domains that require the ability to model complex objects, process large amounts of data, and support complex ad-hoc queries. These new application areas include CAD/CAM, multi-media, data mining, decision-support, and scientific databases. As a result of this expansion, the query execution component is faced with workloads composed of queries that vary significantly in complexity and size. To achieve good system performance for such workloads, it is necessary that resources such as memory, disk bandwidth, and processor bandwidth be effectively allocated during query execution.; In this thesis, we present a new resource management framework for database query execution that is based on concepts from microeconomics. Specifically, we address the difficult problem of managing resources in a multiple-query environment composed of queries with widely varying resource requirements. The central element of the framework is a resource broker that realizes a profit by "selling" resources to competing operators using a performance-based "currency." The guiding principle for brokering resources is profit maximization. In other words, since the currency is derived from the performance objective, the broker can achieve the best performance by making the scheduling and resource allocation decisions that maximize profit. Moreover, the broker employs dynamic techniques and adapts by changing previous allocation decisions while queries are executing.; To determine the viability of the framework, a prototype broker was designed and implemented to manage memory in a multi-user environment. A new dynamic hash join algorithm was developed to allow the broker to dynamically adjust resource allocations. A simulation study demonstrated the effectiveness of the new algorithm compared to previous adaptable hash join algorithms. After evaluating the prototype memory broker against existing memory allocation techniques, the broker was extended to allocate disk bandwidth in addition to memory, and the hash join algorithm was extended to dynamically adapt to changes in its disk bandwidth allocation. The performance studies demonstrate the viability of the broker framework as well as the effectiveness of the adaptable hash join algorithm and the query admission and resource allocation policies developed for the prototype broker. |
