geoknow: making the web an exploratory place for spatial data
TRANSCRIPT
EU-FP7 LOD2 Project Overview . 02.09.2010 . Page 1 http://lod2.eu
Creating Knowledge out of Interlinked Data
http://geoknow.eu
GeoKnowMaking the Web an Exploratory Place for
Spatial Data
Collaborative Project 2012-2015in Information and Communication Technologies
Project No. 318159Start Date 01/12/2012
Presentation by Orri Erling, OpenLink Softwareat
Copernicus Big Data Workshop, Brussels, March 2014
EU-FP7 GeoKnow. 21.01.2013 . Page 2 http://geoknow.eu
Making the Web an Exploratory Place for Geospatial Data
Consortium
Institute for Applied InformaticsGermany
OpenLink SoftwareUnited Kingdom
BroxGermany
OntosSwitzerland
UnisterGermany
Athena Research andInnovation CenterGreece
EU-FP7 GeoKnow. 21.01.2013 . Page 3 http://geoknow.eu
Making the Web an Exploratory Place for Geospatial Data
The Love Story
Ms. Globe dreams of Mr. Cube• Desiloization • Global semantics• Integrated metadata
Mr. Cube dreams of Ms. Globe...• Critical mass?• Global scope• Relevant applications
EU-FP7 GeoKnow. 21.01.2013 . Page 4 http://geoknow.eu
Making the Web an Exploratory Place for Geospatial Data
Mind the Gap !
• Scalability, On-Line Performance?• Data preparation, visualization tool chain?• Cost of deployment, availability of expertise?• Functionality — e.g., reference systems, linestring, polygon,
multipolygon geo collection ...?• Complex functions... contains/within/intersects, intersection,
centroid, convex hull ...?
EU-FP7 GeoKnow. 21.01.2013 . Page 5 http://geoknow.eu
Making the Web an Exploratory Place for Geospatial Data
How will Mr. Cube prove himself ?
Mr. Cube is an adventurer ... He sails to retrieve the Golden Fleece:
• Column store, vectored execution • Scale-out, linear performance gain on clusters• Adaptive schema, i.e., pay for RDF flexibility only when using it
In short, bring the cutting edge of relational technology to GIS and RDF
EU-FP7 GeoKnow. 21.01.2013 . Page 6 http://geoknow.eu
Making the Web an Exploratory Place for Geospatial Data
State of Play
• SPARQL — Virtuoso SPARQL 1.9 G point, 13bn triples LGD fragment, 2 node cluster• SQL single — Virtuoso SQL, 1.3G point OSM, 1 server• SQL Cluster — Virtuoso SQL Cluster, 1.3G point OSM, 2 node cluster• PostGIS — PostgreSQL, 1.3 G point OSM
Workload: Drilldowns zooming down from Europe-wide to city-wide, LOD2 GeoBench
EU-FP7 GeoKnow. 21.01.2013 . Page 7 http://geoknow.eu
Making the Web an Exploratory Place for Geospatial Data
Science Without End• Data Clustering
• GIS mostly accesses nearby objects together• If also stored together, up to 3x boost for both SQL and SPARQL• With a cluster, co-locating geometry and thematic data pays;
obvious with RDB, special trick with RDF
• Adaptive Schema• With RDF, the properties that near-always occur together
may as well be stored physically together• RDF will run just as fast as relational, but retain flexible schema
EU-FP7 GeoKnow. 21.01.2013 . Page 8 http://geoknow.eu
Making the Web an Exploratory Place for Geospatial Data
GeoKnow Generator Stack
• Virtuoso — high performance column store • Instance matching (LIMES, Fagi)• Format conversion, ETL, Enrichment (GeoLift, Fagi-tr)• Visualization (Facete, Mapify, Cubeviz)
EU-FP7 GeoKnow. 21.01.2013 . Page 9 http://geoknow.eu
Making the Web an Exploratory Place for Geospatial Data
Proving The Point (Use Cases)• Benchmarks, cooperation with LDBC for more DI and DB benchmarking
• Unister applies GeoKnow to E-commerce, e.g., tourism
• Brox applies GeoKnow to car manufacturing logistics
EU-FP7 GeoKnow. 21.01.2013 . Page 10 http://geoknow.eu
Making the Web an Exploratory Place for Geospatial Data
Consummation of the Relationship …
Right now, PostGIS is good enough for many GIS applications • Use RDF in Virtuoso: Win in performance, Gain RDF flexibility • Use SQL in Virtuoso and win more.
Later, during GeoKnow• Win 3x more with either SQL or SPARQL• Close the RDF to SQL performance gap
The obstacles to RDF adoption went away ...