Dynamic remote host classification in grid computing using Clonalg

In our work, we study the efficacy of classifying registered hosts dynamically using Clonalg to classify registered hosts during job scheduling. Grid is evolving as the computing structure of the future. The success in commercial grid computing is the ability to negotiate resource sharing arrangements with a set of registered participating parties. Grid computing is capable of integrating services across distributed heterogeneous[1] disparate resources with a centralized control to provide quality of service. Commercial grid computing systems need to process terabytes / petabytes of data or need huge computational power to process business / scientific data. In a commercial environment assigning processing task to remote hosts whose processor power and memory is not being currently used is crucial to improve quality of service. The selection of a remote host based on available processor power or memory is application dependent. Resource discovery algorithms are available but identifying ideal resource to reduce queue time and response time. Experimental results show an improvement of nine percent in data classification using CLONALG over normal methods.