Grid-based utility middleware infrastructure for distributed services provisioning
Loading...
Date
2010
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
As Service Oriented Computing becomes the prevalent methodology for
engineering distributed applications, service composition remains a key
challenge to the reality of on-the-fly application composition using available
services in an ecosystem. Current state-of-the-art service composition
strategies are based on a predefined sequence of actions or workflow and are
unsuitable when applied in dynamic and open settings due to their lack of
scalability, fault-tolerance, and high bandwidth cost overhead. To address this
problem, this thesis makes the case for a new flexible and adaptive strategy
for service composition that is suitable for highly dynamic and open
distributed services provisioning environments.
To support this thesis and demonstrate our solution approach, we
propose MINDS – a Middleware Infrastructure for Distributed Service
Provisioning as a platform to analyze and elaborate our service composition
strategy. MINDS is based on model of business processes as a collaborative
conversation among software agents; It employs late (runtime) binding to
services and facilitates richer interactivity of services by enabling services to
be active and aware of changes in the user and/or execution environment. We
further formulate a new service provisioning life-cycle process for runtime,
automated composition based on MINDS strategy. The new life-cycle process
is not based on a predefined sequence of actions or workflow but rather on
composition goal defined based on user request. To evaluate our strategy and
demonstrate its utility and applicability, we implement an experimental
prototype for e-Tourism Virtual Enterprise Collaboration as a case study. We
also carry out empirical analysis and simulation experiments to compare the
performance of MINDS with related strategies using scalability, fault-tolerance
and bandwidth cost optimization as performance metrics.
The results of the simulation experiments shows that MINDS is scalable
as it require lower service matchmaking and composition time when the
number of services and tasks increase respectively. Moreover, in a sample of
80 tasks, 34 (representing 42.5%) met deadline when the centralized service
agent used in existing strategy failed, while 55 (representing 68.75%) met
deadline when decentralized service agents used in MINDS failed. Also, more
tasks met deadline in MINDS when the service provider failed than the
centralized service agent approach because the failure was discovered at a
minimal time. Finally, the simulation results also indicated that MINDS
optimize bandwidth by requiring less amount of data traffic on the network. In
a sample of 41 participating services in composition, the overhead cost for
data transmissions in both decentralized and centralized scheme were
respectively, 136 Rands and 267 Rands.
We conclude that in view of the limitations of existing strategies for
service composition in dynamic and open settings, a new flexible and adaptive
middleware strategy as proposed in this thesis becomes imperative. Such a
solution offers a more efficient composition strategy in terms of scalability,
fault-tolerance and communication cost. It provides the crucial platform
required in the actualization of the future “Internet of Services” towards onthe-
fly composition and runtime binding of services.
Description
A thesis submitted in fulfillment of the requirement for the degree of Doctor of Philosophy in the Department of Computer Science, Faculty of Science and Agriculture at the University of Zululand, South Africa, 2010.
Keywords
Service Oriented Computing