Scalable management and data mining using astrolabe

Abstract

Astrolabe is a new kind of peer-to-peer system implementing a hierarchical distributed database abstraction. Although deigned for scalable management and data mining, the system can also support wide-area multicast and offers powerful aggregation mechanisms that permit applications to build customized virtual databases by extracting and summarizing data located throughout a large network. In contrast to other peer-to-peer systems, the Astrolabe hierarchy is purely an abstraction constructed by running our protocol on the participating hosts - there are no servers, and the system doesn’t superimpose a specialized routing infrastructure or employ a DHT. This paper focuses on wide-area implementation challenges. © 2002 Springer-Verlag Berlin Heidelberg.