Cyclic query processing in object-oriented databases

Abstract

This paper investigates cyclic query processing issues in objectoriented databases. We first define a data and cyclic query model for an object-oriented database system using a graph model. We then discuss the efficient processing of a general object-oriented cyclic query. For efficient processing, we first define a general cyclic query, the access plans generated for a given query, then develop a cost model to determine the cost for each access plan generated. We also investigate the retrieval algorithms used for actual data retrieval. [i] INTRODUCTIONObject-oriented databases have recently become a very active research area. Several projects on object-oriented databases are mentioned in [DITT86]. Object-oriented database systems research are attempts to overcome the rather limited modelling capabilities that are provided by traditional database systems. These traditional systems are not well-suited to handling advanced applications like CAD/CAM design, office automation, VLSI design and AI applications among others. Object-oriented database systems have the capabilities to handle the complex applications mentioned above by supporting complex data types, generalization/specialization abstraction and aggregation abstraction among others.This paper investigates query processing issues in object-oriented databases. Specifically we are interested in efficient processing of cyclic queries. Query processing issues in an object-oriented database bring out many of the same issues as in conventional databases. One issue in objectoriented query processing is the order in which classes involved in a query must be evaluated. This is similar to the order in which the relations involved in a join must be evaluated in relational databases and the order in which record @pes are visited in evaluating a query in network databases.Query processing involves the following three elements.Identify the class of cyclic queries to consider. Identify all access plans for a given query. Estimate the cost of each access plan and select reasonable access plans for a given query.(1) (2) (3)Most research in cyclic query processing has focused on recursive query processing and optimization in relational and logic databases. In general, the solution to a recursive query cannot be expressed as a finite relational algebraic expression and therefore cannot be evaluated directly by a conventional relational database system. One approach to solve this problem is to provide an appropriate interface between PROLOG and a relational database system [JARK84]; while others provide some means of recursion to overcome the limitations of relationally complete query languages [KERS84, JARK85.LINN861. The optimization techniques for this kind of recursion can be found in [BANC86, ULLM85, GARD861. The above solutions are for conventional systems using flat tables. LJNN871shows that this is cumbersome for the user as well as for the optimization of recursive queries and proposes a better solution combining the non-first normal form relation model with a recurs...

Path dictionary: a new access method for query processing in object-oriented databases

“…In order for OODBSs to be successful, implementation and performance issues must be addressed. Efficient techniques for query processing and indexing are critical [1], [3], [8], [10], [11], [12]. Although indexing techniques have been proposed to support query processing in OODBSs [4], [5], [13], [18], they in general introduce large storage overhead and maintenance cost.…”

Signature path dictionary for nested object query processing

“…These queries are called nested queries. There are two basic approaches to evaluating a nested query: top-down and bottomup evaluations [5]. The top-down approach traverses the objects starting from an ancestor class to a nested class.…”

“…5 Assume that a query involves only the target class and a predicate class and that the classes on the path have equal probability to be used in a query. The average of the optimal false drop probabilities for signatures corresponding to objects in classes C1, Cz, ..., C, is:…”

On graph based interaction for semantic query languages

“…The proposal presented in this paper for the ER model can be easily extended to any graphical query language supporting the notion of class and relationship, enhancing the friendliness of the way in which the query is expressed and, in some case, also the expressive power of the language itself. As an example, it has been reported by several authors [9,24] the lack of expressive power of the object oriented query language presented in [18] based on the view-based approach; the inclusion in that environment of the hybrid strategy discussed in this paper could easily solve the problem.…”

“…Through this second mechanism the expressive power of the graphical query language reaches the one of the conjunctive queries [8]. To specify a set of adjacent concepts can be seen as the basic building block of the user interaction in building a query, and the way in which the user is allowed to do that in the proposals available in the existing proposals can be reduced to two main paradigms of interaction: path-based (see for instance, [14,13,15,19,20,2]) or view-based (see for instance, [26,5,11,18,12,16,7]). …”