Fast Computation of Reachability Labeling for Large Graphs

• Published in: Advances in Database Technology - EDBT 2006 pp. 961 - 979
• Main Authors: Cheng, Jiefeng, Yu, Jeffrey Xu, Lin, Xuemin, Wang, Haixun, Yu, Philip S.
• Format: Book Chapter Conference Proceeding
• Language: English
• Published: Berlin, Heidelberg Springer Berlin Heidelberg 2006; Springer
• Series: Lecture Notes in Computer Science
• Subjects: Applied sciences; Artificial intelligence; Bipartite Graph; Computer science; control theory; systems; Computer systems and distributed systems. User interface; Directed Acyclic Graph; Directed Graph; Exact sciences and technology; Information systems. Data bases; Large Graph; Memory organisation. Data processing; Pattern recognition. Digital image processing. Computational geometry; Software; Speech and sound recognition and synthesis. Linguistics; Transitive Closure; Computer vision; Database query; Very large databases; Information browsing; Reachability; Semantic web; Query processing; Traffic management; Transportation network; Database; Internet; Bioinformatics; Directed graph
• ISBN: 3540329609; 9783540329602
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/11687238_56

The need of processing graph reachability queries stems from many applications that manage complex data as graphs. The applications include transportation network, Internet traffic analyzing, Web navigation, semantic web, chemical informatics and bio-informatics systems, and computer vision. A graph reachability query, as one of the primary tasks, is to find whether two given data objects, u and v, are related in any ways in a large and complex dataset. Formally, the query is about to find if v is reachable from u in a directed graph which is large in size. In this paper, we focus ourselves on building a reachability labeling for a large directed graph, in order to process reachability queries efficiently. Such a labeling needs to be minimized in size for the efficiency of answering the queries, and needs to be computed fast for the efficiency of constructing such a labeling. As such a labeling, 2-hop cover was proposed for arbitrary graphs with theoretical bounds on both the construction cost and the size of the resulting labeling. However, in practice, as reported, the construction cost of 2-hop cover is very high even with super power machines. In this paper, we propose a novel geometry-based algorithm which computes high-quality 2-hop cover fast. Our experimental results verify the effectiveness of our techniques over large real and synthetic graph datasets.