A multi-paradigm approach to query processing: integrating structural, behavioral, and historical knowledge
Anthony Waisanen · University Microfilms International eBooks · 1992
This dissertation describes the general problem of integrating structural, behavioral, and historical knowledge sources in terms of the specific problem of processing access queries. These queries are presumed to be submitted to a federation of semantically heterogeneous, physically distributed, autonomous database systems. The processing of these queries is controlled by a centralized processor. The component databases in the federation provide the centralized processor with structural and behavioral knowledge, views of their respective schemas and statistical profiles (such as the number of tuples expected to be retrieved during a SELECT operation, the cardinality of each relation fragment, the size of each attribute in each relation fragment, and the number of distinct values for each attribute in each relation fragment). These views and profiles, however, may be inaccurate at a given moment since the component databases are not tightly coupled with the federation-level query processing system. Thus, an approach which generates access strategies solely on the basis of schemas and statistical profiles risks the possibility of making erroneous calculations. In our approach, access strategies are generated by using multiple paradigms which cooperate throughout the reasoning process. The views of the data dictionaries and statistical profiles of the component databases are used in conjunction with histories of previously executed queries and query fragments. We show how our approach of using multiple knowledge sources and multiple, cooperating systems avoids regenerating inefficient or costly access strategies even in the presence of incomplete or inaccurate information. An architecture for the system that uses this approach is described in detail. Various approaches to query processing are presented and are contrasted with our approach. We also show how our approach may be applied to a related problem, fault diagnosis in long-distance, telecommunications networks.