INFORMATIONQUALITYEVALUATIONFORGRIDINFORMATION
SERVICES
Wei Xing, Oscar Corcho, Carole Goble
School of Computer Science
University of Manchester
United Kingdom
wxing@cs.man.ac.uk
ocorcho@cs.man.ac.uk
carole@cs.man.ac.uk
Marios Dikaiakos
Department of Computer Science
University of Cyprus, Cyprus
mdd@cs.ucy.ac.cy
Abstract The quality of the information provided by information services deployed in the EGEE production
testbed differs from one system to another. Under the same conditions, the answers provided for
the same query by different information services can be different. Developers of these services and
of other services that are based on them must be aware of this fact and understand the capabilities
and limitations of each information service in order to make appropriate decisions about which and
how to use a specific information service. This paper proposes an evaluation framework for these
information services and uses it to evaluate two deployed information services (BDII and RGMA)
and one prototype that is under development (ActOn).
Keywords: Grid, Grid Information Service, Information Quality, Evaluation

21. Introduction and Motivation
Information Services are regarded as a vital component of the Grid infrastructure. They
address the challenging problems of the discovery and continuous monitoring of all types
of Grid resources, including services, hardware, software and other entities. The quality of
information provided by information systems affects the performance and the behaviour of
other dependent Grid services. For instance, a Grid meta-scheduling service will not work
optimally if the quality of the information used for decision making is poor; a Grid Resource
Broker depends heavily on the quality of the information about Grid resources provided by the
information services that it uses; etc.
There is currently little work done on the evaluation of information quality of Grid infor-
mation services. Most evaluation studies focus on performance measurement [1–2], such as
evaluating scalability, overload, query response time, etc. Such measurements are based on
the assumption that information quality is equal for different information services. However,
this assumption does not hold in reality, since each information system has different mecha-
nisms for collecting and processing information, and adopts different information models for
storage and querying. This is something covered in our experiments, which show that even for
a simple query different systems provide extremely different results. For example, we submit-
ted the query “find me Computing Elements which support the Biomed Virtual Organisation”
simultaneously to the two EGEE default information services, BDII and RGMA, and they had
different results: BDII returned 151 and RGMA returned only 30. Independently of the reasons
for such variable results, the main outcome from this simple test is that information quality of
currently-deployed Grid information services has to be considered carefully.
The work described in this paper has several objectives. First, we want to obtain a fair sys-
tematic approach to measure information quality of different Grid information services,
so that we can compare them and provide guidelines related to the circumstances in which each
of them can be used. The main challenge here is related to the fact that different Grid informa-
tion services have different information models to represent the same type of Grid resources:
some of them use LDAP to represent that information and others use relational models, and
the information that they store about each resource may also differ. Unlike information quality
evaluation in other domains (such asWeb search, where precision and recall measurements can
be obtained by counting numbers of documents), the information objects in our evaluation are
heterogeneous, both in the information model used and in its access API, what makes it hard
to compare the outputs. We have proposed the use of a common information model to allow
comparisons between these outputs. We explain the details in Section 3.2.
Another challenge to be overcome is related to the differences in the querying capabilities
and expressiveness supported by each service, what makes it difficult to design a good set of
relevant experiments for the evaluation. Some services allow making complex queries that
relate information from different domains (computing elements that support a specific virtual
organisation and a specific software environment) and others just provide simple querying
functionalities. In our approach we have proposed a set of representative queries that may be
issued by other middleware services or applications that use these information services, and
which have increasing levels of complexity.
Our second objective is to use the proposed approach for the information quality evaluation
of two information services (BDII and RGMA) deployed in the EGEE Grid and one
prototype that is under development (ActOn). We will analyse the results obtained from
this evaluation and identify the reasons for obtaining such results. We think that these results
can be used by developers working on these Grid information services, in order to improve
them, and by developer of systems that are based on them.
The remaining of this paper is organised as follows. Section 2 describes briefly the infor-
mation systems to be evaluated. Section 3 introduces our evaluation framework, including the
design rationale, the experiments to be carried out, and the metrics to be used for the evaluation,
together with details about how they are measured for each system. Section 4 describes the

Information Quality Evaluation 3
GRIS GRIS
GIIS
Client Client
Information Providers Information Providers
BDII
(a) BDII
GRI SCl i R
eI ntCf i R
oi rmtaR P vavdRvntsi R
? aI Ri ? ? I l vamI n
? I I ? C? ? ? I l vamI n
oP ? ?
e ? mi na
? nsI Rf vamI n? GRI ? mSi Rt
BDI ? ?
(b) RGMA
Figure 1. Overview of the BDII and RGMA architectures
results of the experiments carried out, and provides some conclusions related to these results.
Finally, Section 5 reflects about the lessons learnt in the design of this evaluation framework
and gives references to additional performance tests that we have carried out.
2. Grid Information Services
Currently, there are several well-known and widely-used Grid information services: Mon-
itoring and Discovery System (MDS), Berkeley DB Information Index (BDII), and RGMA
[3–5]. These services are deployed in most Grid systems, such as Europe Data Grid, Cross-
grid, NASA Grid, and Open Science Grid [6–10], and widely used by Grid middleware and
applications running on them. From these three services, we will select BDII and RGMA
for our evaluation, since they are the default information services for the EGEE Grid. We
do not include MDS in our evaluation because it is not deployed as the information service
used for Computing Elements (CEs) and Sites in EGEE and would make difficult to perform
the comparison. Besides, BDII is based on MDS, with the same information model (infor-
mation representation and access), hence the general results regarding information quality and
recommendations obtained for BDII could be easily extrapolated to MDS.
Besides these two services, we will evaluate the ActOn-based information service [11], an
ontology-based information service recently developed by our group, which integrates infor-
mation from both of these services. In fact, we started the development of this service when
we realised that the quality of the information provided by the existing information services
was not good enough, as we will show in Section 4.
In this section, we describe these three information services in detail. Table 3 summarises
some of their main features, including their information model, information access protocol,
and other aspects related to their architecture.
2.1 Berkeley DB Information Index (BDII)
BDII [4] is an improvement of MDS [3] , the information service component of the Globus
platform. It uses the MDS information model and access API and caches information with the
Berkeley DB. In its current version, MDS2.x, information about Grid resources is extracted
by "information providers", which are software programs that collect and organise information
from individual Grid entities, either by executing local operations or by contacting third-party
information sources (e.g., the Network Weather Service, SNMP, etc.).
Extracted information is organised according to the LDAP (Lightweight Directory Access
Protocol) data model, in LDIF format, and uploaded into LDAP-based servers of the Grid
Resource Information Service (GRIS), as shown in Figure 1(a). GRIS servers can register
themselves in the Grid Index Information Services (GIIS) in order to aggregate directories,
using a soft-state registration protocol called Grid Registration Protocol (GRRP).

4An update process is used to populate LDAP-based servers. It consists in obtaining LDIF,
either by doing an ldapsearch on LDAP URLs or by running a local script that generates LDIF.
Then the LDIF is inserted into the LDAP database.
Currently, there are around 200 BDII servers (site BDII plus regional BDII) deployed in the
EGEE production testbed.
2.2 Relational Grid Monitoring Architecture (RGMA)
RGMA [5] is a framework that combines monitoring and information services based on a
relational model, which is implemented with XML. It has been built in the context of the EU
DataGrid project and implements the Grid Monitoring Architecture (GMA) proposed by the
Open Grid Forum.
As shown in figure 1(b), GMA models the information infrastructure of the Grid using three
core types of components: (i) producers, which provide information; (ii) consumers, which
request information; and (iii) a single registry, which mediates the communication between
producers and consumers.
RGMA implements two additional properties over GMA. First, consumers and producers
handle the registry in a transparent way; thus, anyone using RGMA to supply or receive
information does not need to knowabout the registry. And second, all the information appears as
one large relational database and can be queried as such (anyway, in the current implementation,
the database is centralised). RGMA can be accessed using the RGMA API.
In the EGEE production testbed, there are 110 sites, for each of which there is a monitoring
node (MON) that has an RGMA server. All of them use a centralised registry server located at
lcgic01.gridpp.rl.ac.uk.
2.3 Active Ontology (ActOn)-based information service
ActOn [11] is an ontology-based information integration system that can be used to generate
and maintain up-to-date information for a dynamic, large-scale distributed system, such as a
Grid system. It has been developed by the Information Management Group at the University
of Manchester. The ActOn architecture is comprised of a set of knowledge components,
which represent knowledge from the application domain (e.g., the EGEE Grid) and from the
information sources (e.g., RGMA and BDII servers); and software components, such as a
metadata scheduler (MSch), an information source selector (ISS), a metadata cache (MC), and
a set of information wrappers. Figure 2 shows how these components are interrelated and how
they are related to the corresponding information sources where data is taken from.
Two distinct features of the ActOn system, when compared to other information integration
systems, are: 1) it has an intermediate information source selection step that takes into account
the current information needs and the state of the information services to be accessed; 2) and
it also includes a “metadata cache” that works with an update-on-demand policy, which avoids
continuous update requests by aggregating only the metadata that is being queried.
The service that we will evaluate is a deployment of the ActOn system that uses BDII and
RGMA as information sources, and a Grid Ontology [12–13] as its information model, and
has been deployed in the EGEE certificate and production testbeds.
3. An evaluation framework for information quality in Grid
information services
Information quality (IQ) can be defined as ameasure of the value of the information provided
by an information system to its users [14]. There are many characterisations of what quality
means in this context (taking into account that quality is normally subjective and depends on the

Information Quality Evaluation 5
Metadata Scheduler
Distri
buted
Infor
matio
n Sou
rcesDGA
S
Infomation Source Selector
Mediator
W ra p
p er RGMA
InfoSource Ontology
Domain Ontology
Events
1
2 3
4
5
6
7
8
9
BDII
Metadata Cache
<<uses>>10
Figure 2. Overview of the Active Ontology architecture
BDIIRGMA
ActOn-based Information service
Tree
Relational
Graph-based
LDIF
XML
RDF/RDFS/OWL SPARQL Yes
Yes No
No No
Yes
Data Model Representation language Query Language
LDAP
SQL
Architecture
Centralized
Distributed
Centralized
Semantic-able Metadata Cache Information Access
On-demand
On-demand +Data Warehouse
Data Warehouse
Figure 3. Features of the most common Grid information services
intended use of the information by users). The authors in [14] distinguish between intrinsic,
contextual, representational and accessibility IQ, and define different factors to be considered
for each of them (accuracy, objectivity, reputation, relevancy, etc.).
The authors in [15–16] propose to focus on seven of these characteristics, which are con-
sidered the most important ones, independently of their domain: completeness, accuracy,
provenance, conformance to expectations, logical consistency and coherence, timeliness, and
accessibility. In our framework we have selected three of these features, namely completeness,
accuracy and conformance to expectations.
We are not worried about the provenance of information, since we know clearly which are
the information sources that we use in each moment and which are the information providers
responsible for that information. We are not worried either about accessibility, since we as-
sume that the systems work within a Grid security infrastructure (e.g., GSI [17]), so that the
information is accessible as long as the client has the corresponding rights to access it and
knows the information model and API used by the corresponding information service.
With respect to the logical consistency and coherence and the timeliness of the information
retrieved and aggregated from the information sources, these are features that will form part of
our future evaluation work, and will be also considered in further developments of the ActOn-
based information service. An example of why the first feature is important is the following:
there are many cases where a computing element specifies that it gives support to MPI but does
not comply with the requirements for running anMPI job, which are that it must be a CE server,
must have an sshd service running on it, must have the libraries mpirun and libmpi.so in
its file system, and must have at least two worker nodes. Information services like BDII or
RGMA only store and provide the information that their information producers give them,
without checking their consistency, hence they provide incorrect information due to this fact.
As an example of the second feature, BDII normally updates the information that has been

6provided by its information sources every five or six minutes, what means that this information
may be already inaccurate when a client requests it. Hence, having metadata about the lifetime
and freshness of information in the information service is important.
In this section we describe our evaluation framework for information quality, including
metrics to be used for the evaluation, the design rationale, and the experiments to be carried
out, together with details about how the metrics are obtained for each system.
3.1 Evaluation metrics
Tocheck the three criteria considered in our framework, we are interested in knowingwhether
all information services obtain the same results when answering the same query, given the same
conditions in the EGEE testbed. We also want to check how many of those answers are correct
and how many of the existing answers are actually retrieved. This also permits us to know
whether the results provided by the services conform to the expectations of the users. To check
this, we have selected two metrics, commonly used in information retrieval: precision and
recall. Below we provide their definitions and the formulae used to calculate them:
Precision: The proportion of relevant information retrieved, out of all the information retrieved.
Precision =
(relevant information) ∩ (retrieved information)
retrieved information
(1)
Recall: The proportion of relevant information that is retrieved, out of all the relevant infor-
mation available.
Recall =
(relevant information) ∩ (retrieved information)
relevant information
(2)
3.2 Experiment setup and design
Wehave designed a set of experiments formeasuring the information quality criteria selected.
Measurements are taken on a real Grid testbed, the EGEE production testbed, which at the time
of the experiments, has gLite 3.0.1 installed as its middleware. The user interfaces used to
access the EGEE Grid are the UI machines at the University of Manchester1, United Kingdom,
and at the Institute of Physics of Belgrade2, Serbia.
To carry out the experiments and record their results, we have developed a set of Java-based
client software and Unix shell scripts, available at the IST OntoGrid project CVS [18].
Thekey aspects uponwhichwecompare different information services are: i) the information
model that each information service adopts; and ii) the expressiveness of its query language.
In order to evaluate these two features, we have proposed six representative queries that cover a
wide range ofGrid systems, includingGrid hardware resources, software resources, middleware
environment, services, applications, etc., and show increasing complexity. These queries can be
normally issued by middleware systems like schedulers, resource brokers or by more complex
applications:
Query 1: Find all the Computing Elements (CEs) that support the BIOMED Virtual
Organisation (VO).
Query 2: Find all the CEs that support the BIOMED VO and have more than 100 CPUs
available.
Query 3: Find all the CEs that support the MPI running environment.
Query 4: Find all the CEs that support the BIOMED VO, have more than 100 CPUs
available, and support the MPI running environment.
1ui.tier2.hep.manchester.ac.uk
2ce.phy.bg.ac.yu

Information Quality Evaluation 7
Query 5: Find all the CEs where GATE (Geant4 Application for Tomographic Emission)
can be run.
Query 6: Find all the CEs that support the BIOMED VO, have more than 100 CPUs
available, and where GATE can be run.
Table 1. An Example of the Query 1 in BDII, RGMA, and ActOn
Information
Service
Query 1
BDII
(LDAP
Search)
ldapsearch -x -H ldap://lcg-bdii.cern.ch:2170
-b mds-vo-name=local,o=grid ’(&(objectClass=GlueVOView)
(GlueVOViewLocalID=biomed))’ GlueCEAccessControlBaseRule
RGMA
(SQL Query) select GlueCEVOViewUniqueID, Value from
GlueCEVOViewAccessControlBaseRule WHERE Value=’VO:biomed’
ActOn
(SPARQL
Query)
PREFIX egeeOnto: <http://www.cs.man.ac.uk/img/ontogrid#>
SELECT ?ceid ?ceID ?VO
WHERE
?ceid egeeOnto:CEUniqueID ?ceID .
?ceid egeeOnto:hasVO ?VO .
OPTIONAL { ?ceid egeeOnto:VO ?ceID .
FILTER ( ?vo = ‘‘biomed’’)}
Each of these six queries has been translated into the query languages of the three information
services. Table 1 shows an example for Query 1. And we use different clients to execute these
queries and extract the results obtained (e.g., ldapsearch for BDII, the gLite RGMA client tools
for RGMA and a Java-based ActOn client for the ActOn-based information service.
Not only queries are different, but also query results are obtained in different manners, due
to the differences in the information models of each service. The result of a BDII query is a
set of LDAP entries, of an RGMA query a set of table rows, and of an ActOn-based query a
set of RDF triples. Figure 4 shows three different ways to show the same Grid resource in the
three services evaluated (i.e., ce02.tier2.hep.manchester.ac.uk, an EGEE Computing Element).
Even if they have different syntax and size, in our experiment we count them as one piece of
information each. That is, we use each “Grid resource” obtained from a query as the basic unit
for counting information, which will be used to calculate precision and recall, as described in
Section 3.3.
3.3 Experimental Results Measurement
The experiment consists in examining the information retrieved for each of the six queries
aforementioned, so as to get their corresponding precision and recall measures.
Precision is easy to determine, since it can be computed manually by looking at the results
obtained from each query. In all cases, we assume binary relevancy of information, that is,
each piece of information retrieved is either relevant or irrelevant for the issued query.
Recall is more difficult to determine, due to the fact that the amount of information available
in the EGEE production testbed changes frequently in these systems and there is no way to

8| ceid | ceID | VO || <http://img.cs.man.ac.uk/ontogrid1234423456> | "ce02.tier2.hep.manchester.ac.uk" | "biomed" |
# biomed, ce02.tier2.hep.manchester.ac.uk:2119/jobmanager-lcgpbs-biomed, UKI-NORTHGRID-MAN-HEP, local, griddn: GlueVOViewLocalID=biomed,GlueCEUniqueID=ce02.tier2.hep.manchester.ac.uk:2119/jobmanager-lcgpbs-biomed,mds-vo-name=UKI-NORTHGRID-MAN-HEP,mds-vo-name=local,o=gridGlueCEAccessControlBaseRule: VO:biomed
+----------------------------------------------------------------------------------+| GlueCEVOViewUniqueID | Value |+---------------------------------------------------------------------------------- +|ce02.tier2.hep.manchester.ac.uk :2119/jobmanager-lcgpbs-biomed/biomed | VO:biomed |
Query results of BDII:
Query results of RGMA:
Query results of ActOn:
Figure 4. Results of BDII, RGMA, and ActOn for the the same Grid resource Computing Element at University
of Manchester (ce02.manchester.ac.uk)
get accurate information about the actual state of the Grid resources that are available without
using the information services that we are evaluating. To get a good approximation that can
be used for our purposes, we execute each query 100 times, with a 4-minute interval between
executions, that is, we monitor the testbed during 400 minutes. Then we use the highest value
obtained from this 100 executions as the total number of relevant information to be used to
calculate recall.
4. Evaluation Results and Conclusions
Tables 2, 3 and 4 provide the precision and recall measurements obtained after the execution
of the experiments described above for the three information services selected: BDII, RGMA
and the ActOn-based information service. The values provided in the tables show the average
of executing the queries 100 times.
Table 2. BDII Recall & Precision Measurement (100 times)
QueryNo. Retrieved Info. Relevant Info. Precision Recall
1 14,999 15,200 1 0.987
2 242,517 19,708 0.082 0.918
3 7174 7300 1 0.983
4 485034 4600 0.010 0.990
5 - - - -
6 - - - -
Table 3. RGMA Recall & Precision Measurement (100 times)
QueryNo. Retrieved Info. Relevant Info. Precision Recall
1 3417 15200 1 0.225
2 6321 6321 1 1
3 6568 7300 1 0.900
4 11245 4914 0.437 0.563
5 - - - -
6 - - - -
Asageneral comment about these results, we canhighlight the fact thatBDII shows in general
poor results with respect to recall and precision, while ActOn and RGMA present better results.
This is mainly related to the repository that BDII uses (LDAP), which is too lightweight and

Information Quality Evaluation 9
Table 4. ActOn Recall & Precision Measurement (100 times)
QueryNo. Retrieved Info. Relevant Info. Precision Recall
1 15200 15200 1 1
2 34100 34100 1 1
3 6568 7300 1 0.900
4 6568 7300 1 0.900
5 24 24 1 0.900
6 6 6 1 1
hence provides weak information process and query capabilities; while RGMA’s is based on
relational databases and ActOn’s is based on RDF, which both have better query capabilities.
Now we will analyse with more detail some of the system behaviours over specific queries,
and derive more conclusions from these values:
BDII has weak query capabilities. Table 2 shows that BDII has extremely bad precision
results for queries 2 and 4, while the results for queries 1 and 3 are excellent. This is related
to its weak query ability, as aforementioned. LDAP-based queries are string-based, and hence
they cannot be used to support queries over numerical values, such as “greater than or lower
than”. If we want to improve this precision value, we need to fetch all the information about
CE CPUs as a string value first (as we have done to get these results), and then post-process
(filter) those results on the client side. RGMA and the ActOn-based information services do
not have that problem, since their query abilities are better.
RGMA is not able to relate information available in different tables. Table 3 shows that
RGMA has bad precision results in query 4. RGMA contains information to solve this query,
but the information comes from two different tables (GlueCE and
GlueSubClusterSoftwareRunTimeEnvironment), and the query language used by RGMA
does not allow making a join of both tables. Hence the situation is similar to the previous
case: this problem can be solve on the client side by post-processing the results that have been
obtained from each separate query.
RGMA is very sensitive to the registering and availability of information providers at
a given point in time. Table 3 shows that RGMA has bad recall results in query 1. This is
because the amount of Computing Element producers that is available during the experiment is
not always stable, due to the fact that either producers were not registered in the RGMA registry
at that specific moment, or that the producers were not configured correctly or available at that
point in time. BDII and the ActOn-based information service are more robust to this, due to
the fact that they store information locally and do not depend on their information providers at
the time of querying.
Some complex queries cannot be answered by one type of information service in isola-
tion. Tables 2 and 3 show that BDII and RGMA can only answer the first four queries. They
cannot answer queries 5 and 6 because their information providers cannot provide enough in-
formation and should be combined. This shows that the ability of BDII and RGMA to share
their data resources is weak. On the other hand, the ActOn-based information service has
the ability to adopt existing information sources as its information providers, and aggregate
information from these information sources to answer such complex queries.
5. Lessons learned
The experience of developing the experiments for information quality measurement and
conducting them on the EGEE Grid testbed has generated several valuable lessons, most of
them related to the fairness of the information quality measurement process, which can be
applicable to other similar types of experiments.
First, it is difficult to find standard domain-independent methods to measure informa-
tion quality in information systems. Hence if we want to design and run an experiment in

10
a specific domain (e.g., Grid information services), we must design it according to that do-
main and the information needs of the information service users (either other applications or
end-users).
Second, different information services use different information models, and usually
provide different expressivity in their query languages or access APIs. This means that a
special effort has to be made in order to define clearly a fair way to perform measurements that
takes into account these differences.
We think that the results that we have presented can be of great help for the developers who
work in the implementation of these and other similar information services, so that they can
use these experiments as a benchmark suite, and for the developers of information-intensive
applications that make use of these services. In [13], there is also data available about the
performance of the three information services for the same set of queries. We do not include
a discussion about those experiments since they are out of the scope of this paper, but the
main summary would be that BDII and the ActOn-based information service are similar with
respect to their response time and RGMA is generally slower than them, due to its information
management architecture.
Acknowledgements
This work is supported by the EU FP6 OntoGrid project (STREP 511513) funded under
the Grid-based Systems for solving complex problems, by the Marie Curie fellowship RSS-
GRID (FP6-2002-Mobility-5-006668), and by the EU FP6 CoreGrid Network of Excellence
(FP6-004265). We also thank Pinar Alper (IMG group), Antun Balaz and Laurence Field
(EGEE porject), Georges Da Costa and Anastasios Gounaris (CoreGridWP2), for their helpful
comments.

References
[1] X. Zhang, J. Freschl, and J. Schopf., “A performance study of monitoring and information services for
distributed systems,” in 12th International SymposiumonHigh-PerformanceDistributedComputing (HPDC-
12 2003). Seattle, WA, USA: IEEE Computer Society, 2003.
[2] X. Zhang and J. Schopf, “Performance analysis of the globus toolkit monitoring and discovery service,
mds2.” in the International Workshop onMiddleware Performance (MP 2004), part of the 23rd International
Performance Computing and Communications Workshop (IPCCC), April 2004.
[3] K. Czajkowski, S. Fitzgerald, I. Foster, and C. Kesselman, “Grid information services for distributed resource
sharing,” in Proceedings of the Tenth IEEE International Symposium on High-Performance Distributed
Computing (HPDC-10). IEEE Press, August 2001.
[4] “Berkeley Database Information Index (BDII),” http://lfield.home.cern.ch/lfield/cgi-
bin/wiki.cgi?area=bdiipage=documentation.
[5] E. W. Team, “EDG RGMA,” www.marianne.in2p3.fr/datagrid/documentation/rgma-guide.pdf.
[6] “European DataGrid,” http://eu-datagrid.web.cern.ch/eu-datagrid/.
[7] J. Marco and et al., “First Prototype of the Crossgrid Testbed,” in Proceedings of First European AcrossGrids
Conference (AXGrids 2003), LNCS 2970. Santiago deCompostela, Spain: Springer-Verlag, 2003, pp. 67–77.
[8] “Enabling Grids for E-sciencE (EGEE),” http://public.eu-egee.org/.
[9] “Globus Toolkit,” http://www.globus.org/toolkit/.
[10] “The Spallation Neutron Source (SNS) project,” http://www.sns.gov/.
[11] W. Xing, O. Corcho, C. Goble, and M. Dikaiakos, “A Grid Information Service based on an Intelligent
Information IntegrationArchitecture,” inEurope SemanticWebConference 2007 (ESWC-2007), 2007, Poster.
[12] M. Parkin, S. van den Burghe, O. Corcho, D. Snelling, and J. Brooke, “The Knowledge of the Grid: A Grid
Ontology,” in Proceedings of the 6th Cracow Grid Workshop, Cracow, Poland, October 2006.
[13] “OntoGrid CVS,” http://www.ontogrid.net/ontogrid/downloads.jsp.
[14] R. Wang and D. Strong, “Beyond Accuracy: What Data Quality Means to Data Consumers,” Management
Information Systems, vol. 12, no. 4, pp. 5–34, 1996.
[15] N. Dushay and D. Hillman, “Analyzing metadata for effective use and re-use.” in the international DCMI
Metadata Conference and Workshop., 2003.
[16] B. Hughes, “Metadata quality evaluation: Experience from the open language archives community.” in
ICADL, 2004, pp. 320–329.
[17] I. Foster, C. Kesselman, G. Tsudik, and S. Tuecke, “A Security Architecture for Computational Grids.” 5th
ACM Conference on Computer and Communications Security Conference, 1998.
[18] “OntoGrid CVS at UoM,” http://rpc314.cs.man.ac.uk.