100 likes | 240 Views
StormRider: Harnessing “Storm†for Social Networks. Social Network Applications. CHANGE IS A CONSTANT !!. How do you keep up with the current trends and important nodes in an ever-changing network??. How do you lookup previous versions of a network?.
E N D
Social Network Applications CHANGE IS A CONSTANT !! How do you keep up with the current trends and important nodes in an ever-changing network?? How do you lookup previous versions of a network? How do you track changes for a certain node??
9/11 Terrorist Network How do you track what changes in his neighborhood ??
Motivation • No current tools available for dynamic network analysis • Almost all current work focuses on a snapshot of a network at a time • No support for continuous queries on a dynamic network • No specific language to query a social network • Can we build a cloud infrastructure to tackle these two problems??
Architecture Application Model-SN Interface Model-SN Storm Interface Add-Topology Query-Topology Analysis-Topology Jena-HBase Interface HBase View Interface Storm Storage Layer View Layer HBase Views HDFS Files Jena-HBase Store HDFS Files
Sample Add Topology for Twitter Storage Twitter-API-Spout Jena-HBase-Bolt Node-centric View Update-Nodes-View-Bolt Count-User-Degree-Bolt Update-Landmarks-View-Bolt Rank-Users-Bolt Merge-Users-Bolt Landmark-Selection Landmark-centric View
Sample Query Topologies for Twitter O/p to Client I/P SPARQL-Query-Spout • Tracking neighborhoods: Use a topology that runs a continuous query on RDF triples in Jena-HBase using a regular expression SPARQL query • SELECT ?x WHERE { ex:John gleen:onPath( “[twitter:Has_Friend]/[twitter:Has_Friend]” ?x ) } • Version Management: Use the concept of RDF reification to add version information for every triple added to the triple store. Then a SPARQL query of the form: • SELECT ?z WHERE { ?x twitter:Timestamp ?y . ?x rdf:subject <ex:John> . ?x rdf:predicate <twitter:Has_Friend> . <ex:John> <twitter:Has_Friend> ?z . FILTER( ?y <= “2012-01-25T21:00:00”^^xsd:dateTime ) } can be used to find all friends of “John” before 9pm 01-25-2012
Sample Analyze Topologies for Twitter Degree-C-Bolt Twitter-User-Spout Closeness-C-Bolt Betweenness-C-Bolt