Storing and Querying RDF Data in Atlas∗
Zoi Kaoudi, Iris Miliaraki, Matoula Magiridou, Antonios Papadakis-Pesaresi
and Manolis Koubarakis
Dept. of Informatics and Telecommunications
National and Kapodistrian University of Athens
Panepistimioupolis, Ilissia
Athens 15784 Greece
{zoi,iris,matoula,apapadak,koubarak}@di.uoa.gr
Keywords
RDF, query processing, Semantic Web, peer-to-peer net-
works, DHT, scalability.
1. INTRODUCTION
In recent years, more and more resources are semantically
annotated, thus generating huge amounts of RDF metadata.
Current centralized RDF repositories lack the required scal-
ability and fault tolerance to deal with this emerging situa-
tion. Therefore, the need for a scalable system that will be
able to scale to millions of RDF triples is becoming preva-
lent. Distributed hash tables (DHTs) is a recent P2P tech-
nology that has been proposed for the scalable and fault-
tolerant storage and querying of RDF data [2, 1]. Since
annotation is by itself a distributed process it ties very well
with the model of work imposed by P2P systems.
In this demo paper, we present Atlas, a P2P system for the
distributed storage and retrieval of RDF data. Atlas is built
on top of the distributed hash table Bamboo1 and supports
pull and push querying scenarios. It inherits all the nice
features of Bamboo (openness, scalability, fault-tolerance,
resistance to high churn rates) [10] and extends Bamboo’s
protocols for storing and querying RDF data. In the On-
toGrid project, Atlas is used to implement the metadata ser-
vice of S-OGSA, a new architecture for the Semantic Grid
[3].
2. ATLAS ARCHITECTURE
Nodes in an Atlas network are organized in an identifier ring
using the Bamboo DHT protocol. Nodes can enter RDF
∗This work is partially funded by FP6/IST project On-
toGrid. The work was performed while the authors were
with the Dept. of Electronic and Computer Engineering,
Technical University of Crete.
1http://bamboo-dht.org/
Demos and Posters of the3rd European Semantic Web Conference
(ESWC 2006), Budva, Montenegro, 11th - 14th June, 2006
Update?Processor?
RQL queries?RDF data?
ATLAS nodes?
ATLAS protocols?
All ATLAS nodes?cooperate to run?the protocols?
insert RDF triples?
receive answers?
pose?one-time queries?
RQL queries?
subscribe with?continuous?queries? receive?notifications?
ATLAS nodes are?organized in an?identifier ring?
Bamboo DHT network layer?
Local Storage?
Subsription?Processor? RQL Parser?Query?Processor?
Figure 1: Atlas architecture
data into the network and pose RQL [6] queries. In the
typical Atlas application that we envision, RDF data will
be used to describe resources owned by network nodes (e.g.,
ontologies or services). Atlas supports two querying scenar-
ios: one-time querying and publish/subscribe. Each time a
node poses an one-time query, the network nodes cooperate
to find RDF data that form the answer to the query. In the
publish/subscribe scenario, a node subscribes with a con-
tinuous query. A continuous query is indexed somewhere in
the network and each time matching RDF data is published,
Atlas nodes cooperate to notify the subscriber. A high level
view of the Atlas architecture is shown in Figure 1.
In the following, we describe the architecture of each node,
which is similar to the architecture of [2] (see Figure 1).
We distinguish between six components in an Atlas node: