Abstract: "Databases model relationships between entities and such relationships can often be represented as edges in directed graphs. Database queries that construct new relationships by joining existing ones typically correspond to various kinds of reachability problems in graphs. We investig...

Ausführliche Beschreibung

Bibliographische Detailangaben

1. Verfasser:	Jakobsson, Hakan (VerfasserIn)
Format:	Buch
Sprache:	English
Veröffentlicht:	Stanford, Calif. 1993
Schriftenreihe:	Stanford University / Computer Science Department: Report STAN CS 1492
Schlagworte:	Database management Hochschulschrift
Zusammenfassung:	Abstract: "Databases model relationships between entities and such relationships can often be represented as edges in directed graphs. Database queries that construct new relationships by joining existing ones typically correspond to various kinds of reachability problems in graphs. We investigate techniques for solving such reachability problems based on the idea of storing intermediary results that are formed during the computation as trees, where the structures of the trees reflect how the intermediary results were derived. The advantage is that when the trees are used in later steps of the computation, their structures can be used to avoid deriving many duplicate or irrelevant results. We give algorithms using tree-based techniques for a number of transitive closure and shortest paths problems We analyze these algorithms using connectivity measures relating to the number of node-disjoint paths between the nodes in the graph. In particular, the pairwise connectivity of two nodes, x and y, denoted CONN(x, y)), is an interesting measure. We show that, unlike previous algorithms, our tree-based transitive closure algorithms have upper bounds that are O([sigma][subscript (x, y)] [element of] v x v CONN (x, y)). We compare our algorithms to several others based on different approaches and show that using trees can lead to great improvements in many cases and, even in the worst case, introduces only a marginal overhead compared to standard techniques. We discuss some factors that determine the actual improvements We also consider the join-ordering problem that arises when several relations are composed and show that a simple cheapest-first heuristic works much better with tree-based compositions than with standard ones.
Beschreibung:	Stanford, Calif., Univ., Diss.
Beschreibung:	VIII, 59 S.