Autonomic Middleware Services for Just-In-Time Grid Services Provisioning

Abstract

The advent of widespread distributed computing environment, such as information systems and computational grids computing has enabled a new generation of applications that are based on seamless access, aggregation and interaction. The dramatic side of the story is a strong presence of the plea that those decentralized Grids are potentially affected by a number of primitives derived from their anatomy, in that, they are inherently large, complex, heterogeneous and dynamic, globally aggregating a large number of independent computing and communication resources. This has clearly exposed an essential exigency for a vital change to how these applications are developed and managed, which has motivated researchers to consider other techniques used by biological systems to deal with such problems. This is referred to as autonomic computing, which is defined by Ganek and Corbi^[1] “… as a collection and integration of technologies that enable the creation of an information technology computing infrastructure for the next era of computing—e-business on demand …”. This study presents a computational model to support just-in-time and on-demand services for autonomic computing. Service reservation and job schedule systems are employed in this model to estimate the required services in advanced. Intelligent classification is utilized to cluster consumers into groups sharing the same behaviour and hence offer the required services for each consumer in advance, according to the group’s usage pattern of application services. To this end, a machine learning middleware service based on Self-Organizing Maps (SOM) is designed, developed and implemented to carry out the intelligent classification for the autonomic computing. A case-study scenario of intelligent connected homes is demonstrated in this study to show the usability of such system.