Metadata Management in S-OGSA
Oscar Corcho, Pinar Alper, Paolo Missier,
Sean Bechhofer, Carole Goble, and Wei Xing
School of Computer Science, University of Manchester
Oxford Road, Manchester M13 9PL, United Kingdom
{ocorcho,penpecip,pmissier,seanb,carole,wxing}@cs.man.ac.uk
Abstract. Metadata-intensive applications pose strong requirements for
metadata management infrastructures, which need to deal with a large
amount of distributed and dynamic metadata. Among the most relevant
requirements we can cite those related to access control and authorisation,
lifecyclemanagement andnotification, anddistribution transparency.This
paper discusses such requirements and proposes a systematic approach to
deal with them in the context of S-OGSA.
1 Introduction
Metadata can be attached to many different types of objects, available in dif-
ferent formats and locations, and can be expressed in a wide range of languages
(e.g., natural language, lists of terms, formal languages) and using a wide range
of vocabularies (e.g., keyword sets, concept taxonomies). Depending on these as-
pects, many authors use the terms semantic versus non-semantic, or rich versus
non-rich metadata.
Many technologies are available to manage metadata, such as Jena, Sesame,
Boca, Oracle-RDF, Annotea, Technorati, etc. They give a good level of support
for current metadata-intensive applications. However, they fall short for some of
the metadata management requirements that new applications are posing, such
as metadata distribution, lifecycle management and authorisation.
This paper starts by providing a set of requirements for advanced metadata
management (Section 2), which characterise metadata-intensive applications in
domains like bioinformatics, social sciences, engineering, market analysis, etc. Ex-
isting approaches and technologies do not meet all of the previous requirements [1].
Hence we make a proposal for managing metadata as first-class resources in dis-
tributed systems, so that we can deal with the previous requirements (Section 3).
Finally we provide some conclusions and future work.
2 Metadata Management Requirements
In this Section we describe some of the advanced requirements for metadata
management that are needed in some metadata-intensive applications.
Metadata should be stored and accessible in a distributed man-
ner. Many systems will need to integrate distributed metadata (and ontologies),
Y. Shi et al. (Eds.): ICCS 2007, Part II, LNCS 4488, pp. 712–719, 2007.
c
© Springer-Verlag Berlin Heidelberg 2007

Metadata Management in S-OGSA 713
which may be supplied by multiple parties and with different technologies. Most
of the current metadata management systems provide repositories designed for
centralized use, with metadata consumers and producers acting as local-access
clients using specialized APIs. Remote distributed access to metadata and on-
tologies has received little attention, causing each system to depend on one
particular technology, thus reducing interoperability.
A service-oriented approach to metadata and ontology access could improve
this situation. However, this should not be done by simply wrapping current
metadata management systems with web services that replicate their local APIs,
as described below.
Metadata should be accessible with technology-independent service-
oriented protocols. Metadata can be available in multiple forms (in RDF, as
social tags, in natural language, etc.). Hence we require systems that enable the
integration and exploitation of this heterogeneous metadata.
Even in the case of a single form of metadata (e.g. RDF(S)) each technology
uses a different abstraction and means to handle it. There is a clear need for a
common abstraction for metadata, which would enable the systematic develop-
ment of services that combine other services and applications with varying levels
of semantic capabilities for dealing with and interpreting metadata.
Metadata should evolve together with the resources that it de-
scribes and the vocabularies that it uses. Metadata evolves due to differ-
ent reasons. However, the dynamics of metadata is poorly supported by existing
technologies, although some work has been done with respect to the semantics
of change and its propagation to the related metadata [2,3].
Metadata should be maintained up-to-date in a cost-effective manner. This
includes maximising the automation of different aspects of the knowledge life-
cycle, managing the evolution and change of metadata and knowledge models
in distributed contexts, and synchronising adequately the evolution of all these
related entities by means of notification mechanisms.
Metadata access should be controlled. Metadata owners may want to
define the conditions and rules under which others can access their metadata.
However, existing technologies vary in their support and granularity for security,
which normally means access-control only. For instance, Sesame provides built-in
user/role-based access control per repository, while Jena has no built-in access-
control support since it assumes that this can be provided by the underlying
database technology. Moreover, all these access control mechanisms have their
own specific APIs and conventions, which creates overhead at the client layer
(need for multiple usernames and passwords, or multiple sign-ons).
Securing message exchanges, authentication and access-control are well-studied
problems in distributed environments, with standards1 and open-source reference
implementations2 for global user identification, single sign-on, communication en-
cryption and representation and decision of resource-sharing policies.
1 E.g., XACML: www.oasis-open.org/committees/xacml/, WS-Security: www.oasis-
open.org/committees/wss/
2 E.g., Globus GSI: http://www.globus.org/toolkit/docs/4.0/security/

714 O. Corcho et al.
3 Metadata Management in S-OGSA
In this Section we describe the approach for metadata management proposed
in the context of the Semantic-OGSA (S-OGSA) architecture [4]. S-OGSA is
an extension of the Open Grid Service Architecture [5] for the development of
distributed applications that need to use explicit and distributed metadata.
The S-OGSA model is driven by the principle that associations between re-
sources and their metadata should be first-class resources that can be distributed
and managed in a service-oriented manner. This gives support to the first two
requirements from Section 2, and is the basis for giving support to the rest.
The term Semantic Binding (SB) [6] denotes this new type of resource, which
constitutes the core of S-OGSA. Semantic Bindings represent associations be-
tween any set of resources and any set of knowledge entities (e.g., ontologies, rule
sets, controlled lists of terms). These associations contain metadata, which can
be encoded in different formats : RDF, natural language, as social tags, etc. Be-
sides, different Semantic Bindings can describe the same set of entities : different
tools or persons may create different annotations for the same resource.
The following subsections describe the capabilities associated to Semantic
Bindings and how they support our metadata management requirements.
3.1 Semantic Binding Capabilities
In [4] we describe the model used to represent Semantic Bindings, expressed as an
ontology that extends the Grid ontology described in [7]. The main properties of a
Semantic Binding are the set of resources to which it refers (that is, the resources
for which it contains metadata), the set of knowledge entities that the metadata
is based on, and the actual metadata that they store. Other properties (state,
creation time, last modification time, etc.) are stored and used for managing
lifetime, notification and authorisation mechanisms.
Figure 1 shows the basic operations provided by the Semantic Binding service
suite (the Semantic Binding Service, its corresponding Factory and a Metadata
Service that unifies the metadata stored by several Semantic Bindings).
– Create. It creates a Semantic Binding, given the described resources, the
Knowledge Entities used for the description, and the metadata to store.
– Update Resource and Knowledge Entity References. They allow managing the
references to Resources and Knowledge Entities of the Semantic Binding.
– Update Semantic Binding Content. It updates the metadata stored in the
Semantic Binding, due to its reannotation or curation.
– Destroy. It destroys the Semantic Binding, together with its content, imme-
diately or at a scheduled point in time.
– Archive. It archives the Semantic Binding content so that it is not active but
can be retrieved in case that it is needed later (e.g., for provenance).
– Query. It executes a query over the metadata stored by the Semantic Bind-
ing. Queries will be sent in a query language that the Semantic Binding
supports, and can take into account the knowledge entities to which the
Semantic Binding refers or not.

Metadata Management in S-OGSA 715
Fig. 1. Functionality of the Semantic Binding Service
This suite is supported by the S-OGSA reference implementation3, which
complies with WS-Resource Framework [8] (WSRF). It can be deployed on the
Globus Toolkit 4 platform4 and in Apache Tomcat5.
Next we describe how we deal with the rest of requirements, namely lifetime
management, metadata change notifications and controlled access to metadata.
3.2 Semantic Binding Lifetime
In some applications, metadata is a dynamic entity subject to frequent changes:
resources and knowledge entitites that they refer to can evolve, become suddenly
unavailable, be destroyed, etc. Some of these changes may cause metadata to
become invalid. Other changes may not have an influence on metadata validity
(e.g., the removal of a concept in an ontology for which there are no instances in
the stored metadata). Finally, metadata may become invalid after a given period
of time, when a new annotation tool has been made available and the resource
has to be reannotated, when a metadata curation process is in place, etc.
In order to deal with all these changes in a principled way, S-OGSA defines
SBs as stateful resources with a defined lifetime and identifies the states and state
transitions that a SB can go through throughout its lifetime. The corresponding
state diagram is presented in the next subsection.
Figure 2 shows the state diagram associated to an SB, which includes fun-
damental states and state transitions, and the external events that cause the
transitions. The SB lifetime specification extends WS-ResourceLifetime, a part
of WSRF that standardizes the way that resources are destroyed, and defines re-
source properties for the inspection and monitoring of a resource lifetime. While
3 Available at the OntoGrid CVS.
4 http://www.globus.org/toolkit/
5 http://tomcat.apache.org/

716 O. Corcho et al.
WS-ResourceLifetime is focused exclusively on resource destruction, we extend
it to include any life-changing event that may affect the validity and updates
of an SB. Furthermore, the basic state machine presented here can be extended
with sub-states if needed in a certain application.
The explanation of the state transition diagram is as follows. When it is first
created, a Semantic Binding SB is in the Valid state. We denote with ResSB
and KESB, respectively, the set of Resources and Knowledge entities that are
part of the association, and with contentSB the metadata payload within SB.
Fig. 2. State transition diagram for a generic Semantic Binding
State transition events are of the following types:
– Changes in the described resources, denoted by ResSB → Res′SB.
– Changes in the Knowledge entities, i.e., KESB → KE′SB
– Updates to the SB content: contentSB → content′SB.
Note the Resources and Knowledge entities can also be destroyed: ResSB → ∅,
ResSB → ∅. In addition to these external events, a content expiration date can
also be associated to an SB, so that it gets stale upon expiration.
For a Valid SB, these events cause its transition to either one of two possible
Validate states, Validate Res and Validate KE. These are interim states in which
the SB may be invalid, and is awaiting re-validation. A re-validation process,
either manual or automated, is any procedure that updates any or all of ResSB,
KESB, or contentSB, and which results in a decision as to whether the updated
entities represent a new valid combination6.
– For a ValidateRes SB, such procedure determines whether the existing meta-
data can be associated to the new Resources, and provides an update to the
references in SB to Res′SB. For example, following a change in a workflow
6 In both cases, the SB goes back to the Valid state in case of successful validation,
and to Invalid otherwise.

Metadata Management in S-OGSA 717
that is described with a piece of metadata, the procedure determines whether
the same metadata can be associated to the new workflow.
– For a ValidateKE SB, the problem is to determine whether the new ontology
can still be used to interpret the old metadata.
Finally, the Archived state indicates that a SB is still available for inspection.
3.3 Notification of Semantic Binding Changes
Requirement 3 suggests that the metadata-aware services that use SBs should be
informed of state change for those SBs. For this purpose, S-OGSA defines a set
of notification mechanisms based on WS-Notification. S-OGSA proposes using
a set of pre-defined topics associated to the changes described above, which any
consumers can subscribe to.
Services that receive any of these notifications will decide, as part of their
business logic, how to react to the changes that they are notified about. The
S-OGSA specification does not enforce any specific type of behaviour. However,
as part of our future work, we are designing a service, called SB housekeeping
service, which monitors SB lifetime by subscribing to all their topics. This service
will be responsible for activating application-dependent re-validation procedures
(validation of the SB content, triggering of re-annotation processed, etc.).
3.4 Security over Semantic Bindings
The last requirement suggests the need to control the access to metadata in more
fine-grained ways than what it is currently done with current technology.
S-OGSA provides the framework needed to support security in metadata man-
agement. Metadata is a first-class resource; hence standard security mechanisms
can be applied to metadata in the same way as it is done with other resources,
including the possibility of specifying and enforcing access control policies over
each SB. Besides, since the reference implementation of S-OGSA can be deployed
on top of Globus Toolkit 4, its associated security mechanisms, such as Globus
GSI, can be also applied, ensuring both message and transport level security,
based on standard X.509 end entity and proxy certificates.
With this framework, it is possible to allow or deny access to the different
annotations of an object, stored by different SBs, based on the users or groups
that have created them and on their access control policies.
4 Conclusions and Future Work
In this paper we have described the requirements posed by some of the existing
and the envisaged metadata-intensive distributed applications:
– Applications may require their explicit metadata to be encoded in multi-
ple forms, supplied by multiple parties and coming from different contexts.
Moreover, metadata may need to be in different physical locations.

718 O. Corcho et al.
– Applications may use heterogeneous metadata storage and query technolo-
gies, each of which has specific advantages over the others (e.g. Sesame is
good at query performance, Jena has a rich API, etc.). To ease metadata
sharing, common technology-independent means for metadata access (meta-
models and protocols) become necessary.
– Metadata normally evolves, either because of new metadata generation
means or because of the evolution of the resources or knowledge entities
that it refers to. Adequate means to deal with this evolution have to be
made available.
– Access to metadata may need to be secured with different levels of granu-
larity and different access control policies.
We have shown how the approach followed in S-OGSA complies with the
previous requirements without the need to create expensive and difficult-to-
maintain ad-hoc solutions for each of them. Basically, S-OGSA proposes to treat
metadata as a first-class resource (Semantic Binding) in the application, so that
standard resource management and sharing solutions used in the context of
distributed applications can be easily applied to it. This includes aspects like
service orientation, lifecycle management, security, etc.
The S-OGSA approach is being used in the development of different types of
applications, all of which are characterised by being metadata-intensive and by
needing support for some of the previous requirements. The S-OGSA reference
implementation is being or will be used in several system prototypes: a satellite
quality image analysis system [9], an information service for the EGEE Grid [10],
and a coral bleaching alert system (Semantic Reef) [11].
Our future work will be devoted to improve our S-OGSA reference implemen-
tation, which includes the Semantic Binding Service suite presented in Section 3.
We will consider the additional requirements that the early adopters of S-OGSA
are providing and create an implementation with industrial standards.
Among the aspects that will be improved, we can cite the following:
– Security. We will provide a set of pre-defined security configurations that
cover the most common metadata management security aspects.
– Naming. Our current implementation uses WS-Addressing EndPoint Refer-
ences (EPRs) to identify Semantic Bindings and the Resources that they
refer to, and URIs to identify Knowledge Entities. However, there are more
ways to identify entities in a distributed environment (URIs, ARK Iden-
tifiers, LSIDs). We will implement a metadata identification model using
WS-Naming, which builds on WS-Addressing and extends it with URIs.
– Semantic Binding Housekeeping Service. We will build a configurable SB
Housekeeping Service that gives support to the most common behaviours
for metadata evolution found in applications.
– Support for more forms of metadata. We will give support to several forms of
annotation (besides RDF), such as social tags, natural language comments,
other ontology languages, etc.

Metadata Management in S-OGSA 719
Finally, as part of our future work we will also evaluate our system with respect
to other metadata management systems, in terms of memory consumption, query
execution performance, etc., in distributed settings.
Acknowledgements
This work is supported by the EU FP6 OntoGrid project (STREP 511513)
funded under the Grid-based Systems for solving complex problems, and by
the Marie Curie fellowship RSSGRID (FP6-2002-Mobility-5-006668). We also
thank the other members of the OntoGrid team at Manchester for their helpful
discussions: Ian Dunlop and Ioannis Kotsiopoulos.
References
1. O. Corcho, P. Missier, P. Alper, S. Bechhofer, and C. Goble, “Principled metadata
management for next generation metadata-intensive systems,” in 4th European
Semantic Web Conference (ESWC2007). Submitted, Innsbruck, Austria, 2007.
2. L. Stojanovic, “Methods and tools ontology evolution,” Ph.D. dissertation, Univ
Karlsruhe (TH), 2004.
3. M. Klein, “Change management for distributed ontologies.” Ph.D. dissertation,
Vrije Universiteit Amsterdam, 2004.
4. O´. Corcho, P. Alper, I. Kotsiopoulos, P. Missier, S. Bechhofer, and C. A. Goble,
“An Overview of S-OGSA: A Reference Semantic Grid Architecture,” Journal Web
Semantic, vol. 4, no. 2, pp. 102–115, 2006.
5. I. Foster, H. Kishimoto, A. Savva, D. Berry, A. Grimshaw, B. Horn, F. Ma-
ciel, F. Siebenlist, R. Subramaniam, J. Treadwell, and J. V. Reich, The
Open Grid Services Architecture, Version 1.5, gfd-i.080 ed., GGF, July 2006,
http://forge.gridforum.org/projects/ogsa-wg.
6. P. Missier, P. Alper, O. Corcho, I. Kotsiopoulos, I. Dunlop, W. Xing, S. Bechhofer,
and C. Goble, “Managing semantic grid metadata in S-OGSA,” in Cracow Grid
Workshop 2006. Submitted, Cracow, Poland, 2006.
7. M. Parkin, S. van den Burghe, O. Corcho, D. Snelling, and J. Brooke, “The Knowl-
edge of the Grid: A Grid Ontology,” in Proceedings of the 6th Cracow Grid Work-
shop, Cracow, Poland, October 2006.
8. K. Czajkowski, D. Ferguson, I. Foster, J. Frey, S. Graham, I. Sedukhin, D. Snelling,
S. Tuecke, and W. Vambenepe, “Web Services Resource Framework (WSRF),”
Globus Alliance and IBM,” Technical report,, March 2005.
9. M. Sa´nchez-Gestido, L. Blanco-Abrun˜a, M. de los Santos Pe´rez-Herna´ndez,
R. Gonza´lez-Cabrero, A. Go´mez-Pe´rez, and O. Corcho, “Complex data-intensive
systems and semantic grid: Applications in satellite missions,” in Proceedings of the
2nd IEEE International Conference on e-Science and Grid Computing (e-Science
2006), Amsterdam, The Netherlands, December 2006.
10. W. Xing, O. Corcho, C. Goble, and M. Dikaiakos, “Active ontology: An information
integration approach for highly dynamic information sources,” in Submitted to
ESWC2007, May 2007.
11. T. S. Myers, “The Semantic Reef: Managing complex knowledge to predict coral
bleaching on the great barrier reef.” in In proceedings of AusGrid 2007, 2007,
http://eprints.jcu.edu.au/1131.