Token-based Predictive Scheduling of Tasks in Cloud Data-centers

Kumar Narander, Saxena Swati
2015 Research Journal of Recent Sciences   unpublished
Resource Management in a utility-based system such as cloud computing requires a careful observation of users demands and availability of resources. For optimal resource provisioning, an effective task/job scheduling is required which must guarantee fair chance to users and profit to service providers along with maximum utilization of resources. This paper presents a token-based scheduling mechanism which lines up tasks to resources in a fair manner based on a user's token value. A user's token
more » ... is characterized by his/her SLA parameters, his/her duration in the task queue and the task's nature, i.e. computation-intensive, memory-intensive or communication-intensive. Further, to ensure optimal usage to cloud's resources, the token-based scheduling is complemented by a predictive scheduling which matches user's demands with resource's supply, and delays a task in case it's demand is not currently fulfilled by a machine by giving preference to another task. The experimental results of the proposed work strengthen our claim of fairness, profitability and effective resource management.