Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
. 
. 
RDFox 
. Optimised RDF Triple Store 
. Parallel Datalog reasoner (OWL 2 RL) 
. Faster than all competitors 
O F O F ...
Upcoming SlideShare
Loading in …5
×

RDFox Poster

696 views

Published on

Abstract:
We present a novel approach to parallel materialisation (i.e.,
fixpoint computation) of datalog programs in centralised,
main-memory, multi-core RDF systems. Our approach comprises an algorithm that evenly distributes the workload to cores, and an RDF indexing data structure that supports efficient, ‘mostly’ lock-free parallel updates. Our empirical evaluation shows that our approach parallelises computation very well: with 16 physical cores, materialisation can be up to 13.9 times faster than with just one core.

Published in: Technology
  • Be the first to comment

  • Be the first to like this

RDFox Poster

  1. 1. . . RDFox . Optimised RDF Triple Store . Parallel Datalog reasoner (OWL 2 RL) . Faster than all competitors O F O F & D E O T S A T R E I P L R T F D R A L O G R E A SO N E R Y T I S R E V I N U O R D X RDFox . . Features . Low memory consumption per triple . Highly scalable w.r.t. CPU threads . Easy integration as Java or C++ library . SPARQL 1.1 querying support . Preliminaries A Knowledge Base (KB) comprises . Data — e.g. RDF triples . Ontology — e.g. Datalog Program or OWL 2 ontology Ontologies relate notions and concepts; used for . Ontology Based Data Access (OBDA) . Enhanced query answering Example Data Hotel(Hilton), B&B(Blue Ox), Hostel(YMCA) Ontology B&B (x) → Accom(x), Hotel (x) → Accom(x), Hostel (x) → Accom(x) ery Accom(?x) ≈ ‘Which accomodations are there?’ Answers Hilton, Blue Ox, YMCA. . Motivation Core service of KBs: ery Answering ! . Boom-up Approach: compute all logical consequences (the materialisation) of the KB beforehand + short query times (no reasoning at query time) – higher memory usage (materialisation has to be stored) Datalog is a rule & query language . Underpins Semantic Web languages (RDFS, pD∗, OWL 2 RL, SWRL) . OWL 2 EL reducible to Datalog . Recursive language . Increasing popularity in Semantic Web Community Modern computer systems have . Multi-core processors . More and more RAM (>1TB) . . Our Novel Solution RDFox has . Optimised indexing of RDF data — fast reasoning/querying . Almost lock-free update of indexes — low contention . Per-fact parallelisation of materialisation — high scalability RDFox supports . Multithreading — exploits modern hardware . RDF — standardised wide spread data format . Datalog — covers a range of OWL fragments . SPARQL 1.1 — standardised versatile query language . Evaluation Shows speedup factor achieved by n threads w.r.t. 1 thread .. 8 . 16 24 32 20 18 16 14 12 10 8 6 4 2 Threads Speedup .. Clar..osL . Clar.osLE . DBpe.diaL . DBpe.diaLE LUBML01K . LUBM.U01K .. 8 . 16 24 32 20 18 16 14 12 10 8 6 4 2 Threads Speedup . UOBM.L01K . UOBM.U010 . LUBM.LE01K . LUBM.L05K . LUBM.LE05K . LUBM.U05K ClarosL ClarosLE DBPediaL DBPediaLE LUBML01K LUBMU01K Triples (mill) 96 555 118 1530 323 219 Mem (GB) 4.2 18.0 6.1 52 9.3 11.1 import (s) 58 367 215 timeKB(1) 2477 4128 158 8457 73 135 timeKB(max n) 127 285 24 602 7 10 UOBML01K UOBMU010 LUBMLE01K LUBML05K LUBMLE05K LUBMU05K Triples (mill) 430 36 333 911 1661 1094 Mem (GB) 20.2 1.1 13.8 49.0 75.5 53.3 import (s) 783 6 415 2999 timeKB(1) 532 2738 947 442 4859 635 timeKB(max n) 31 168 71 39 745 54.3 Performance . Still rises with hyperthreading (>16 cores) . Almost linear speed-up up-to 8 threads — flaens due to bus conjestion . Outruns all competitors [1] Publication with technical details and in depth information: [1 ] Materialisation of Datalog Programs in Centralised, Main-Memory RDF Systems, Motik et al., AAAI2014. . . Current Work & Outlook Active Development . Constantly developed and maintained . Parallel reasoning with owl:sameAs added . Incremental reasoning implemented Future work . Parallelising incremental reasoning . Improving query processing . Using RDFox in distributed computer networks Visit www.rdfox.org Boris Motik, Yavor Nenov, Robert Piro, Ian Horrocks forname.lastname@cs.ox.ac.uk

×