MODELLING FOR QUALITY OF SERVICES IN DISTRIBUTED GEOPROCESSING
Richard ONCHAGA
International Institute for Geo-Information Science and Earth Observation (ITC)
Enschede, The Netherlands
onchaga(@itc.nl
KEY WORDS: Distributed GIS, web service, Quality, Performance, Internet, Spatial Infrastructures, quality of service
ABSTRACT
Geographic information (GI) services are increasingly the state of the art in geoprocessing, promising novel geospatial
solutions, products and services. Dynamic chaining of GI services however, presents exciting quality of service (QoS)
challenges. Consequently, sustainable use of GI services demands careful consideration of the diverse user requirements
and the complexities of dynamic service chaining in view of the QoS limitations of the Internet. An initial and necessary
step towards provision of QoS guarantees in Internet GIS is QoS-aware service chaining, where services are discovered
and composed on the basis of their non-functional quality characteristics. The paper presents an architecture for QoS-
aware service chaining and defines an extensible QoS model for GI services. Further, to facilitate the successful evolution
of GI enterprise into effective service providing nodes on the GI service infrastructure, an integration framework for GI
enterprises presented. While the provision of hard QoS guarantees on the Internet remains elusive, QoS-aware service
chaining will greatly improve the quality experienced by Internet GI applications and users.
1 INTRODUCTION
The thrust of GIS research in recent years has centered
on web-based geographic information (GI) services. Gl
services are modular components of geospatial computing
applications which are self-contained, self-describing and
can be published, located and invoked across a network
to access and process geospatial data from a variety of
sources (Doyle and Reed, 2001). Built to open interfaces,
the services can be discovered and chained at run-time to
deliver functionality of value to a user . Consequently, GI
services enable cross-system and cross-organisation inter-
operability, promising a whole range of novel applications
as they emerge as the state of the art in geoprocessing.
Services that are dynamically chained present a new model
for delivering geospatial solutions, products and services
to diverse communities of users. Chaining of services is
achieved via one or combinations of the following three
architectural patterns; user defined (transparent) chaining,
workflow-managed (translucent) chaining, or aggregate ser-
vices (opaque chaining) (ISO/TC211, 2002). The major
difference between the architectural patterns is the level
of control the user has over the chaining process. On the
one extreme, in transparent chaining, the user defines and
controls the order and execution of the services. On the
other extreme, in opaque chaining, the user has little or no
control on the chaining and execution of services. Clearly,
the different patterns are meant for different types of users
- transparent chaining is for highly skilled users whereas
opaque chaining is for naive users. Translucent chaining
falls between these extremes.
Existing implementations of GI services are rather simple
and static, with limited capability for dynamic chaining. In
a bid to address the above limitations, current GI industry
efforts are focussed on migrating Gl service specifications
to comply with general XML-based web service standards
of SOAP, WSDL, and UDDI (OGC, 2003). Services based
on the new specifications can be discovered dynamically
D
and chained using established standards like BPELAWS
(Andrews et al., 2003).
It is anticipated that as the technology becomes mundane,
GI services will proliferate. Similarly, the number of users
of these services will explode. Furthermore, ubiquitous GI
services will enable exciting but quality of service (QoS)
sensitive applications like mobile and wireless location-
based services. Meanwhile, services will get increasingly
integrated in mission critical business processes. As a re-
sult, quality of service (QoS) emerges as a critical concern
and a major factor of competition.
QoS is a prerequisite to an effective infrastructure for GI
services, and by extension, a dynamic and sustainable GI
market. Providing QoS guarantees on the Web remains
a big challenge because of the best-effort service model
of the Internet, limitations of the messaging and transport
protocols, and heterogeneity of the underlying distributed
resource platform (Spreitzer and Janssen, 2000, Wang, 2001,
Mani and Nagarajan, 2003). The large volume and high
compute-intensity characteristics of spatial data further ac-
centuate the QoS problem.
Sustainable exploitation of GI services therefore demands
careful consideration of diverse use requirements and QoS
related constraints on the complex process of dynamic ser-
vice chaining and the unique character of spatial data in
view ofthe limitations of Web computing. Moreover, users
will have different QoS requirements which the composed
service chain should meet.
Typically, a service chain comprises disparate services that
are discovered and orchestrated at run-time. Individual
services have distinct non-functional quality characteris-
tics like performance, cost, reliability, etc. Furthermore,
the number and availability of services will vary over time,
as new services become available and others are retired.
A first step towards providing QoS guarantees in service
chains will thus be to enforce QoS-aware discovery and
composition of services, whereby services are selected from
Internat
a comr
on thei
the ser
sequen
workfk
specific
Furthei
predice
in whic
How €:
ners in
an opel
The pa
ery and
model :
functio
chains.
framew
based a
The fra
for des
Gl ente
frastruc
2 TH
Over re
(Alame
an evol
tions, p
user re
with se
produci
The GI
that col
the nod
busines
and BP
other, n
laborati
a marke
vice inl
comme
3 QO
Quality
plines «
some d
like ‘qu
CUSLONT
theme i
tion of
tives.
In this
tive me