Querying Large-scale RDF Datasets Using the SANSA Framework

Abstract

In this paper we present Sparklify: a scalable software component for e cient evaluation of SPARQL queries over distributed RDF datasets. In particular, we demonstrate a W3C SPARQL endpoint powered by our SANSA framework's RDF partitioning system and Apache Spark for querying the DBpedia knowledge base. This work is motivated by the lack of Big Data SPARQL systems that are capable of exposing large-scale heterogeneous RDF datasets via a Web SPARQL endpoint.

Introduction

One of such systems, built on the concepts of RDDs and SparkSQL, is Sparklify [ 6 ] { a scalable software component for e cient evaluation of SPARQL queries over distributed RDF datasets, already integrated into the SANSA framework. It evaluates SPARQL queries while transforming them into a lower-level of Spark programs. Our Sparklify query processor interfaces with Apache Spark and can be used programmatically within Big Data work ow construction and in SPARQL server mode. In contrast, most of the existing systems only feature cli-oriented prototypes.

In this demonstration, we present and describe our implementation of the SANSA SPARQL endpoint. This interface allows access to the RDF data using the common SPARQL endpoint which provides an easy-to-use method to run SPARQL queries via the Web2.

This is an accompanying poster paper for Sparklify [ 6 ], which was accepted at the ISWC resource track. The addition made in this demo is the SPARQL endpoint Web interface for Sparklify, which enables executing SPARQL queries on-the- y using the Web interface i.e., without engineering e orts and command line interfaces. 2

Querying large-scale RDF data: DBpedia as a use case DBpedia [ 3 ] is among the largest and most well-known sources of structured information on the Web. The public DBpedia SPARQL endpoint3 at present o ers access to 438.336.350 triples representing information in a variety of domains. In addition, with public downloads of auxiliary information the data size exceeds a billion triples. Due to its size and variety, DBpedia is often used as a test-bed for novel Linked Data technology in order to analyze strengths and weaknesses. For these reasons, we choose to run Sparklify over the DBpedia knowledge base. The whole pipeline is described below.

Figure 1 depicts the Sparklify architecture. The data (e.g. DBpedia knowledge graph) rst has to be loaded on a distributed le system (Step 1 ). In our case, we use Hadoop Distributed File System (HDFS) for storing the RDF graph that SANSA can read e ciently. Afterwards, data ingestion is performed (Step 2 ). SANSA reads RDF data into an initial RDD of triples. Sparklify applies data partition in parallel over the initial RDD (Step 3, part one). This partitioning facilitates fast querying by giving both the SPARQL-to-SQL rewriter and the Spark processor additional opportunities for pruning during static query analysis. These partitioned data are then queried, based on the input SPARQL query (Step 1 ) using the Sparq lify system { a SPARQL to SQL rewriter. This leverage the potential of the optimizers of both the rewriter as well as those of the underlying frameworks for SQL. The raw output of a query execution is a pair comprised of the resulting RDD together with a mapping for construction of the corresponding SPARQL result set, i.e. a set of bindings. The mapping associates each of the requested SPARQL query's result variables with an expression over 2 Demo. resources: https://github.com/SANSA-Stack/SANSA-Examples 3 https://dbpedia.org/sparql SANSA Engine 3

Partitioning Data Ingestion

RDF Layer

Views Views Sparklifying Query Layer

4 Distributed Data

Structures the RDD schema, such that from each RDD row a corresponding binding can be computed. Hence, the output is again a distributed data structure which can further processed in a distributed fashion (Step 4 ) or visualized or as a result set of records (Step 5 ). Figure 2 demonstrates reuse of a third-party SPARQL-based Wikidata visualization app4 for DBpedia data served with SANSA. 3

Conclusion

Processing and querying RDF data becomes challenging when the size of the data increases. To solve it, many systems have been proposed and try to solve it using the distributed computing framework. Most of the existing systems mostly are cli-oriented prototypes. This brings many confusion and a lot of engineering e ort in order to run SPARQL queries over it. Most of the users, who are familiar with the semantic web technologies, are capable of writing SPARQL queries. But, when it requires cluster con guration, and running prototypes from the source; it is considered as a dead-end. Therefore, we wanted to bring Sparklify { a scalable software framework for e cient evaluation of SPARQL queries over distributed RDF datasets. It contains a user interface as a SPARQL endpoint for easy to write and query RDF data distributed across the Spark cluster.

Acknowledgment

This work was partly supported by the EU Horizon2020 projects BigDataOcean (GA no. 732310), Boost4.0 (GA no. 780732), SLIPO (GA no. 731581) and QROWD (GA no. 723088). 4 https://github.com/stevenliuyi/wikidata-visualization

References

1. Armbrust , M. , Xin , R.S. , Lian , C. , Huai , Y. , Liu , D. , Bradley , J.K. , Meng , X. , Kaftan , T. , Franklin , M.J. , Ghodsi , A. , Zaharia , M. : Spark SQL : Relational Data Processing in Spark . In: Proceedings of the 2015 ACM SIGMOD International Conference on Management of Data . pp. 1383 { 1394 . SIGMOD '15, ACM , New York, NY, USA ( 2015 ). https://doi.org/10.1145/2723372.2742797, http://doi.acm. org/ 10 .1145/2723372.2742797

2. Graux , D. , Jachiet , L. , Geneves , P. , Layada, N.: SPARQLGX: E cient Distributed Evaluation of SPARQL with Apache Spark . In: Groth, P. , Simperl , E. , Gray , A. , Sabou , M. , Krotzsch, M. , Lecue , F. , Flock, F. , Gil , Y . (eds.) The Semantic Web { ISWC 2016 . pp. 80 { 87 . Springer International Publishing, Cham ( 2016 )

3. Lehmann , J. , Isele , R. , Jakob , M. , Jentzsch , A. , Kontokostas , D. , Mendes , P.N. , Hellmann , S. , Morsey , M., van Kleef, P. , Auer , S. , Bizer , C. : DBpedia - a large-scale, multilingual knowledge base extracted from wikipedia . Semantic Web Journal 6 ( 2 ), 167 { 195 ( 2015 ), http://jens-lehmann.org/files/2014/swj_dbpedia.pdf

4. Lehmann , J. , Sejdiu , G. , Buhmann, L., Westphal , P. , Stadler , C. , Ermilov , I. , Bin , S. , Chakraborty , N. , Saleem , M. , Ngonga Ngomo , A.C. , Jabeen , H.: Distributed semantic analytics using the SANSA stack . In: ISWC Resources Track ( 2017 )

5. Schatzle, A. , Przyjaciel-Zablocki , M. , Skilevic , S. , Lausen , G.: S2RDF: RDF querying with SPARQL on Spark . Proc. VLDB Endow . 9 ( 10 ), 804 {815 (Jun 2016 )

6. Stadler , C. , Sejdiu , G. , Graux , D. , Lehmann , J.: Sparklify: A Scalable Software Component for E cient Evaluation of SPARQL Queries over Distributed RDF Datasets . In: Proceedings of 18th International Semantic Web Conference ( 2019 )

7. Zaharia , M. , Chowdhury , M. , Das , T. , Dave , A. , Ma , J., McCauley , M. , Franklin , M.J. , Shenker , S. , Stoica , I. : Resilient distributed datasets: A fault-tolerant abstraction for in-memory cluster computing . In: Proceedings of the 9th USENIX conference on Networked Systems Design and Implementation . pp. 2 { 2 . USENIX Association ( 2012 )