Agent-Based Infrastructure for Dynamic Composition of Grid Services

Introduction

The Service Oriented Computing (SOC) paradigm is currently a promising methodology for engineering applications in both scientific and industrial communities (Bhatia, 2005). It promises a world of distributed services that are loosely coupled to flexibly create dynamic business processes and agile applications that may span many organizations and computing platforms.

A major assumption of the SOC approach is that services can be combined together in order to obtain new, composite services. The implication of this is that future enterprise and scientific applications do not have to be developed from the scratch but assembled on the fly from an ecosystem of services available on a network such as the grid (NGG Report, 2006). This is already having revolutionary impact on the way software applications are designed, architected, delivered and consumed.

Building scientific or business applications that utilize distributed services is one of the challenges facing the SOC research community today. Traditional software engineering methodologies are often inadequate in addressing the complexity and dynamic nature of large-scale distributed systems. Classical techniques for composing services are procedural and only define how different services ought to be invoked e.g. in terms of ordering and parallelism among them. They are geared for low-level invocation of services, and not specially geared for enabling composition (Singh & Huhns, 2005). Moreover, a number of existing techniques are too rigid and are controlled by a centralized composition engine. They are, therefore, inadequate for service composition, especially, when services require capabilities of autonomy, flexibility and robustness.

Interestingly, there has been an explosion of interest in peer-to-peer (P2P) methodology as a promising approach for engineering large scale distributed systems (Bhatia, 2005). P2P systems generally, offer the advantages of scalability, reliability, and robustness over traditional server-based systems (Verma, 2004). Multi-agent Systems (MAS) have thus, emerged as a P2P methodology that address the issues of organizing large-scale software systems and could, therefore, complement existing methodologies for SOC. Agents are autonomous, computational entities that can be viewed as perceiving their environment through sensors and acting upon them through effectors (Bradshaw, 1997). Many existing and potential industrial applications of multi-agent systems in engineering large-scale systems are found in many fields such as e-commerce, telecommunications, transportation, sensor network, scheduling, and manufacturing (Weiss, 1999). Multi-agent systems would, therefore, be invaluable in emerging SOC environments especially in developing and analyzing intricate models and theories of interactivity and collaboration among Collaborative Virtual Enterprises (CVEs) and Virtual Organizations (VOs). Moreover, agents provide greater flexibility and robustness in how services are used and created by operating rationally in a variety of environmental circumstances and context that change in a usually unforeseen manner.

We, therefore, present in this chapter an approach that augment current service-based technologies with multi-agent system for dynamic composition of grid services.

The rest of the chapter is organized as follows: