A Cloud-Based, Geospatial Linked Data Management System - Transactions on Large-Scale-Data-and Knowledge-Centered Systems XX - page 64

Yars2 [ 7 ] is a federated semantic search engine for performing interactive
query answering over heterogeneous LD collected from many disparate Web
sources. The local indexing scheme adopted comprises: (a) keyword indices based
on Apache Lucene to enable keyword lookups, (b) full quad sparse indices, and
(c) join indices to speed up queries. For global-based indexing, three partitioning
methods are employed to decide on the node where a particular quad will be
indexed.
Mika and Tummarello [ 14 ] have produced a research prototype in the form
of a back-end for the Sesame Triple Store which exploits Pig to load and query
RDF data, where RDF loading is performed by converting RDF to Pig's data
model.
Tanimura et al. [ 22 ] have implemented a scalable RDF data processing frame-
work which exploits parallel database processing over the Google File System
(GFS). Hadoop is used as the basic infrastructure based on GFS and MapReduce
while Pig is used as the data processing platform. For ecient RDF querying,
a particular RDF storage scheme which combines vertical partitioning with the
Hadoops key-value data format was adopted.
Husain et al. [ 9 ] have developed a scalable and fault-tolerant framework which
exploits a particular scheme for storing RDF Data in the Hadoop File System
and supports data intensive query processing.
A RDF storage and querying prototype system has been implemented in [ 21 ]
based on MapReduce and HBase. The realized storage scheme employs six HBase
tables to cover all RDF triple pattern combinations, while triples are indexed
through the HBase index structure on row key.
A distributed RDF prototype store is presented in [ 17 ] based on MapReduce
and HBase. The storage scheme employs three indices to cover particular triple
pattern combinations stored in HBase tables in the form of key-value pairs.
Franke et al. [ 4 ] have implemented a prototype with two different distributed
RDF storage schemes based on HBase and MySQL Cluster, respectively. The
HBase database schema relies on creating two tables for storing RDF triples,
while the MySQL-based scheme relies on a simple table which has as columns
the triple subjects, predicates and objects, respectively.
An extension of the RAPID prototype system is proposed in [ 18 ] which relies
on Pig and Hadoop and exploits PigLatin as the high-level language to support
ad-hoc processing and querying over large data-sets.
An RDF molecule-based store has been realized in [ 16 ] by exploiting Hadoop
to scale-out the distributed query processing. A number of extensions with
respect to molecule hierarchy and structure are proposed to the initial mole-
cule definition to resolve particular query performance issues.
Proprietary Approaches. Dydra 6 is a multi-tenant, cloud-based graph data-
base deployed on the Amazon Cloud, which exhibits various features, such as
versioning and disaster recovery. RDF data are stored as a property graph which
directly represents the relationships between them.
6 www.dydra.com .