The Query Web: Reliable Indexing In Dynamic Networks

Peer-To-Peer (P2P) networks have become popular with computer science researchers because of their compelling properties like robustness, scalability and failure resistance. Due to the lack of centralized control, several challenges arise in P2P networks: (1) fast indexing of data structures to allow efficient query processing, (2) creation, abstraction and dissemination of suitable metadata, (3) dynamic assignment of roles to peers according to their capabilities and (4) maintenance of global values in P2P networks. We develop a novel architecture for query processing which we term “The Query Web” to address these challenges within a dynamic scenario where nodes regularly join and leave the network. By combining a distributed hash table approach with unstructured P2P networks, we expect the index to be reasonably insensitive to both changes of the network structure as well as updates of the data instances. This is a major improvement over other systems which usually trade efficient reliable lookups against the ability to cope with change.