Lehrstuhl Fur Informatik Iii’s scientific contributions

Not only in deductive databases, logic programming, and constraint satisfaction problems but also in object bases where each single dot in a path expression corresponds to a join, the optimizer is faced with the problem of ordering large numbers of joins. This might explain the renewed interest in the join ordering problem. Although many join ordering techniques have been invented and benchmarked over the last years, little is known on the actual effectiveness of the developed methods and the cases where they are bound to fail. The problem attacked is the discovery of parameters and their qualitative influence on the complexity of single problem instances and on the effectiveness of join ordering techniques including search procedures, heuristics, and probabilistic algorithms. Thus an extensive analysis of the search space is carried out, with particular emphasis on the existence of phase transitions in this space and on the influence the parameters have on these transitions.

The main idea underlying the paper is to extend the relational algebra such that it becomes possible to process queries against the World-Wide Web. These extensions are minor in that we tried to keep them at the domain level. Additionally to the known domains (int, bool, float, string), we introduce three new domains to deal with URLs, html-documents or fragments thereof, and path expressions. Over these domains we define several functions that are accessible from the algebra within the subscripts of the relational operators. The approach allows us to reuse the operators of the relational algebra without major modifications. Indead, the only extension necessary is the introduction of a map operator. Further, two modifications to the scan and the indexscan are necessary. Finally, the indexscan which has the functionality of a typical meta-search engine is capable of computing a unified rank based on the tuple order provided by the underlying search engines.

The field of object-oriented databases is suffering from the lack of a---widely agreed upon---formal model. In this paper we develop the basics of a formal model for an object-oriented data model (kernel). The formalization is based on dynamic logic: combining many-sorted algebras with states and state transitions. The many-sorted algebras are utilized to model the static aspects, i.e., the value-oriented part of the object model. The state transitions account for the dynamic behavior of the object base due to object modifications. The main dynamic features that we incorporate in the formal model comprise (object) attribute assignment, object creation, object deletion, and set object manipulation, i.e., removing and inserting elements. The virtues of the proposed model are demonstrated on three sample applications. Dangling references may lead a given program to nonterminate ---we will show this fact. Second, a desirable property of a program will be verified. Last, the semantics of a ...

In this paper we address the problem of optimizing the evaluation of boolean expressions in the context of object-oriented data modelling. We develop a new heuristic for optimizing the evaluation sequence of boolean expressions based on selectivity and cost estimates of the terms constituting the boolean expression. The quality and efficiency of the heuristic is evaluated based on a quantitative analysis which compares our heuristic with the optimal, but infeasible algorithm and other known methods. The heuristic is based on the selectivity and evaluation-cost estimates of the terms of which the boolean expression is composed. Deriving these inputs of the heuristics, i.e., the selectivity and cost estimates, is then addressed. We use an adaptation of well-known sampling for estimating the selectivity of terms. The cost estimation is much more complex than in the relational context due to the possibility of invoking functions within a boolean expression. We develop the decapsulation met...

Current data models like the NF 2 model and object-oriented models support setvalued attributes. Hence, it becomes possible to have join predicates based on set comparison. This paper introduces and evaluates several main memory algorithms to evaluate efficiently this kind of join. More specifically, we concentrate on the set equality and the subset predicates. 1 Introduction Since the invention of relational database systems, tremendous effort has been undertaken in order to develop efficient join algorithms. Starting from a simple nested-loop join algorithm, the first improvement was the introduction of the merge join [1]. Later, the hash join [2, 7] and its improvements [19, 22, 28, 39] became alternatives to the merge join. (For overviews see [27, 37] and for a comparison between the sort-merge and hash joins see [13, 14].) A lot of effort has also been spent on parallelizing join algorithms based on sorting [10, 25, 26, 34] and hashing [6, 12, 36]. Another important research ar...