194
G. E. Modoni and M. Sacco
the semantic interoperability of a technological system, i.e. the latter’s capability
to exchange information with other systems [17]. In this regard, the SWT provide
a valid support to the Linked Open Data (LOD) [3], i.e. the paradigm more and
more used for connecting data to be published both on and off the Web. Another
key advantage of SWT adoption consists in their capability to support (some form
of) reasoning, which allow to entail and infer new meaningful knowledge about the
already defined concepts and their linking relationships. Through a concept map,
main aspects related with RDF stores are reported in Fig. 1.
The Resource Description Framework (RDF) is the W3C standard model for data
interchange on the basis of SWT [19]. It offers a flexible and domain-agnostic syntax
to abstract the information as lists of statements which are in turn represented under
the form of triples consisting of a subject, a predicate, and an object. In this regard,
one of the RDF strengths is that it allows to express virtually any type of information,
without the need to previously adapt its data reference structure as in the relational
schema-based approach. As the flip side, the expressivity and flexibility poses the
need to review, for information represented in RDF syntax, traditional data management issues such as efficient persistence, query processing and optimization. In
particular, the fine grain of the RDF data models (triples instead of whole records)
increases the number of the joins included in the queries, making not trivial the formulation of complex queries and also introducing several issues of scalability. The
databases for the persistence and management of any type pf data specified in RDF
syntax are termed RDF stores (also known as Triple stores). Their main components
are the repository and the Application Programming Interface (API), which communicates with the underlying repository to programmatically expose the main database
services. The first solutions of RDF stores have been implemented leveraging existing databases which are in turn based on the more traditional and widely proven
relational model. However, despite this approach allowed to build working solutions
with little effort, the well-defined and rigid structure of the relational model on which
this approach is based does not fit well with a flexible model such RDF [25]. For this
reason, the current efforts of researchers and technicians are addressed towards the
implementation of so called native RDF stores, i.e. purpose-built databases which,
not depending on rigid schemas, fits more properly to the flexible structure of RDF
data.
Native RDF stores are usually considered representatives of the NoSql databases,
whose market is large and not homogeneous. Their most well-known classification
groups existing implementations in terms of the following categories: (a) Column
Databases: e.g. Cassandra, HBase, etc.; (b) Document Databases: e.g. MarkLogic,
MongoDB, etc.; (c) Key-value Databases: e.g. Project Voldemort, Dynamo, etc.; (d)
Graph Databases: e.g. Neo4J, AllegroGraph, etc. [8, 11]. This last category comprises
the native RDF stores, since RDF data can be thought in terms of a directed labeled
graph, where each graph’s node can represent the subject or object of a triple, while
the arc is the predicate that links subject and object. Moreover, it should be noted that,
in addition to the native triple stores, other solutions (e.g. “native” graph databases
or belonging to other NoSql category) are available to handle RDF data, even they
are not designed mainly for this purpose [7].
G. E. Modoni and M. Sacco
the semantic interoperability of a technological system, i.e. the latter’s capability
to exchange information with other systems [17]. In this regard, the SWT provide
a valid support to the Linked Open Data (LOD) [3], i.e. the paradigm more and
more used for connecting data to be published both on and off the Web. Another
key advantage of SWT adoption consists in their capability to support (some form
of) reasoning, which allow to entail and infer new meaningful knowledge about the
already defined concepts and their linking relationships. Through a concept map,
main aspects related with RDF stores are reported in Fig. 1.
The Resource Description Framework (RDF) is the W3C standard model for data
interchange on the basis of SWT [19]. It offers a flexible and domain-agnostic syntax
to abstract the information as lists of statements which are in turn represented under
the form of triples consisting of a subject, a predicate, and an object. In this regard,
one of the RDF strengths is that it allows to express virtually any type of information,
without the need to previously adapt its data reference structure as in the relational
schema-based approach. As the flip side, the expressivity and flexibility poses the
need to review, for information represented in RDF syntax, traditional data management issues such as efficient persistence, query processing and optimization. In
particular, the fine grain of the RDF data models (triples instead of whole records)
increases the number of the joins included in the queries, making not trivial the formulation of complex queries and also introducing several issues of scalability. The
databases for the persistence and management of any type pf data specified in RDF
syntax are termed RDF stores (also known as Triple stores). Their main components
are the repository and the Application Programming Interface (API), which communicates with the underlying repository to programmatically expose the main database
services. The first solutions of RDF stores have been implemented leveraging existing databases which are in turn based on the more traditional and widely proven
relational model. However, despite this approach allowed to build working solutions
with little effort, the well-defined and rigid structure of the relational model on which
this approach is based does not fit well with a flexible model such RDF [25]. For this
reason, the current efforts of researchers and technicians are addressed towards the
implementation of so called native RDF stores, i.e. purpose-built databases which,
not depending on rigid schemas, fits more properly to the flexible structure of RDF
data.
Native RDF stores are usually considered representatives of the NoSql databases,
whose market is large and not homogeneous. Their most well-known classification
groups existing implementations in terms of the following categories: (a) Column
Databases: e.g. Cassandra, HBase, etc.; (b) Document Databases: e.g. MarkLogic,
MongoDB, etc.; (c) Key-value Databases: e.g. Project Voldemort, Dynamo, etc.; (d)
Graph Databases: e.g. Neo4J, AllegroGraph, etc. [8, 11]. This last category comprises
the native RDF stores, since RDF data can be thought in terms of a directed labeled
graph, where each graph’s node can represent the subject or object of a triple, while
the arc is the predicate that links subject and object. Moreover, it should be noted that,
in addition to the native triple stores, other solutions (e.g. “native” graph databases
or belonging to other NoSql category) are available to handle RDF data, even they
are not designed mainly for this purpose [7].