Univerza na Primorskem Fakulteta za matematiko, naravoslovje in informacijske tehnologije

četrtek, 12. november 2015 Ponedeljkov SEMINAR računalništva in informatike

V ponedeljek, 16. novembra 2015, bo ob 16.00 uri v prostorih Fakultete za matematiko, naravoslovje in informacijske tehnologije Univerze na Primorskem, Glagoljaška 8, Koper predavanje v okviru PONEDELJKOVEGA SEMINARJA RAČUNALNIŠTVA IN INFORMATIKE Oddelkov za Informacijske znanosti in tehnologije UP FAMNIT in UP IAM.

PROSTOR: FAMNIT-1-MP2 ob 16:00

SPEAKER: Iztok Savnik

TITLE: big3store: large-scale distributed triple-store (part 2)


Huge graphs are recently stored and managed in triple-store database systems, where each arc is represented by triple: (subject,predicate,object). The scalability of storage system and query processing for managing from Tera towards Peta triples is currently possible by using large-scale distribution of data into shared-nothing clusters.

Query execution system in such environment must be able to employ various types of parallelism to allow simultaneous execution of queries and, therefore, provide reasonable response time. Distributed triple-store big3store is based on dataflow architecture of query processing. Each query is a tree of algebra operations that is dynamically mapped to the tree composed of processes interconnected by streams of graphs. The scheduler that maps query trees to the set of processes balances the computation load among the servers of cluster.

While in teh 1st part of this seminar: 1) recent appearance of graph databases (knowledge graphs and linked data) on the Web, and 2) short overview of graph database management systems, were presented, this 2nd part will focus on the 3) the architecture of big3store and research problems involved in its design and implementation. The development of big3store system is a joint project between Yahoo Japan Research