Database Reference
In-Depth Information
RDF-3X [
15
] is a complete, open-source RDF management system based
on three key principles: (a) physical design is workload independent, (b) query
processor is RISC-style relying on merge joins, and (c) the query optimizer
focuses on join order for the generation of plans. RDF-3X stores all triples in a
clustered B+-tree. RDF-3X models 2D spatial data through WKT and indexes
them via R-Trees.
YAGO2 [
8
]
12
is an extension of the YAGO Knowledge Base enabling the
modeling of entities, facts and events according to the additional aspects of time
and space. Time is represented via the
yagoDate
data type typically with a
resolution of days or sometimes in years. All entities with a permanent spatial
extent on Earth are subsumed by the
yagoGeoEntity
class. The position of such
entities is described via geographical coordinates expressed via a special data
type called
yagoGeoCoordinates
which is a pair of latitude and longitude values.
As this is quite restricting for locations with a physical extent, particular rules
are exploited in order to assign a specific position for particular location types,
such as cities or military and industrial establishments. Special types of spatial
relations are exploited to express a permanent location of an entity, to associate
facts with the location they concern, and to express relations between time and
location.
TDB
13
is part of Jena for RDF storage and querying and can be used as a
high performance store on a single machine which can be accessed either through
scripts or programmatically via the Jena API. The indexing scheme of TDB com-
prises: (a) a node table (dictionary) storing the representation of RDF terms,
(b) three full triple and quad indexes for unnamed and named graphs, respec-
tively and (c) the prefixes table which uses a node table and a index for GPU.
3 Background
In our previous work [
10
], we have developed a similar system to the one being
proposed in this article for which the basic components were analysed. The
added-value and advancement of this article lies on the fact that a better system
architecture has been adopted which takes into account the current and prospec-
tive usage of the system and is able to scale appropriately in such a way that
the query performance remains at satisfactory levels while the updating of LD
is carried out in a lazy fashion by considering the fact that such updating does
not jeopardise the integrity of the application running on top of the system as
well as does not deteriorate the system's performance. In addition, the exposed
functionality of the system has been extended to also handle geospatial LD as
well as different publishing and exporting ways. Moreover, in comparison to pre-
vious work, the system has been heavily tested and evaluated to indicate that it
exhibits the desired performance under the most extreme situations.
Search WWH ::
Custom Search