TMManaging Complex and Varied Data with the IndexFabric 1,2 1,2 1,3 1 1 1Neal Sample , Brian Cooper , Michael Franklin , Gísli Hjaltason , Moshe Shadmon and Levy Cohen 1 2 3RightOrder Inc. Department of Computer Science Computer Science Division 3850 N. First St. Stanford University University of California San Jose, CA 95134 USA Stanford, CA 94305 USA Berkeley, CA 94720 USA {nsample,cooperb}@db.stanford.edu, franklin@cs.berkeley.edu,{gislih,moshes,levyc}@rightorder.com the product, information about the manufacturer, Abstract information about vendors, and so on. Relational databases store different entities separately, and then use Emerging networked applications present significant joins to reconstruct relationships among them at query challenges for traditional data management techniques time. The reconstruction of complex objects such as for two reasons. First, they are based on data encoded in XML documents or LDAP hierarchies can require XML, LDAP directories, etc. that typically have complex numerous “self-joins”. This reconstruction process is inter-relationships. Second, the dynamic nature of typically quite expensive, and limits the performance of networked applications and the need to integrate data existing systems. from multiple sources results in data that is semi- or The IndexFabric solves this problem by maintaining irregularly structured. The IndexFabric has been data relationships explicitly in an elegant, ...