Abstract:
Local resources available at a node are often insufficient to solve large computing newlineproblems. At the same time, underutilized resources remain unused newlinebecause of ignorance of their capabilities, or incompatible administrative restrictions. newlineTo preserve the investment in equipment, and allow solving large newlinecomputational problems, mechanisms are needed to join these independent newlinesystems into cooperating groups across the boundaries of administrative domains newlineand physical proximity. newlineThis cooperation is named as distributed computing that has many flavors like, newlineCloud computing, Grid computing, and Cluster computing. These distributed newlinecomputing fields are concerned about aggregation of distributed computing newlinepower for solving large-scale problems in science, engineering, and commerce. newlineHowever, application composition, resource management, and scheduling in newlinethese environments are complex undertakings. This is due to the geographic newlinedistribution of resources that are often owned by different organizations having newlinedifferent usage policies. newlineDue to the aggregation of heterogeneous resources, resource management is newlineessential for Grid computing. This makes resource management in Grid systems newlinedistinct from traditional computation platform. Therefore, most task newlinescheduling algorithms developed for traditional platforms are not applicable newlineto Grid systems. Resource management includes searching, selecting, scheduling, newlineand monitoring. This thesis focuses on scheduling aspect of Grid computing newlineresource management while job submission, execution, and monitoring newlineare delegated to user and provider middleware. newlineEfficiency of scheduling algorithms affects the user and service provider. newlineEffectiveness of a scheduling algorithm is measured using response time, newlinemakespan, cost, deadline, budget, and communication overhead. A Grid newlinescheduling algorithm is employed at two levels - local scheduling and global newlinescheduling. Local scheduling algorithms manage the nodes within site and newlineimprove the system performance, while global scheduling algorithms select.
