Atlas: Storing, updating and querying RDF(S) data on top of DHTs

Transactions on Large-Scale Data- and Knowledge-Centered Systems XIII

“…The results of the earlier version of this paper [10] have influenced most of our work on P2P computing over the last years, inspiring us to develop IF functionality [27] in the Minerva system [53], study IF in an XML context [69], design DHT-based digital libraries [52], and implement Web/Grid service registries [70] for the EU projects OntoGrid and SemsorGrid4Env.The deployment of these ideas on various domains demonstrates the generality of the problem and shows that our protocols may be applied beyond the adopted scenario. The results of the earlier version of this paper [10] have influenced most of our work on P2P computing over the last years, inspiring us to develop IF functionality [27] in the Minerva system [53], study IF in an XML context [69], design DHT-based digital libraries [52], and implement Web/Grid service registries [70] for the EU projects OntoGrid and SemsorGrid4Env.The deployment of these ideas on various domains demonstrates the generality of the problem and shows that our protocols may be applied beyond the adopted scenario.…”

Storing and Querying Semantic Data in the Cloud

“…The disadvantage is that usually no compute node knows how many compute nodes exist and how the data is distributed among all compute nodes. Usually in systems like Edutella [37,97], RDFPeers [25], PAGE [35], GridVine [6,31], RDFCube [87], Atlas [67], 3RDF [12], [13] and [101], a distributed index is used that routes requests to the compute node storing the requested data. Since this distributed index is the central component of the architecture, the implementation choice of the index determines how the triples of the RDF graph are assigned to the compute nodes.…”

“…Ring structure. In RDFPeers [25], PAGE [35], Atlas [67] and [82] the compute nodes are order, e.g., by their IP address. This order defines the direct neighbours of each compute node.…”

“…In RDFPeers [25], GridVine [6,31], Atlas [67] and 3RDF [12] a hash cover is applied which each triple is stored on at most three compute nodes based on the hash of its subject, property and object. As a consequence all triples in which one resource occurs are located on the same compute node.…”

SPARQL Query Optimization on Top of DHTs

“…Motivated by [1,11], our group has been developing Atlas (http://atlas.di.uoa.gr), a full-blown open source P2P system for the distributed processing of RDF(S) data stored on top of DHTs. The RDFS reasoning functionality, the architecture and various applications of Atlas are presented in [7][8][9].…”