An Enhanced MinMin Resource Scheduling Algorithm for Scientific Applications in Cloud Computing

Abstract

Cloud computing is a very well known term which is used commonly in almost every business or research field. It is a young but familiar technology which enables the clients to use its services without being bothered to know how the services run and leaving this job to the providers. Cloud providers offer three main services namely IaaS, PaaS and SaaS where IaaS stands for Infrastructure as a Service, PaaS stands for Platform as a Service and SaaS stands for Software as a Service. A user can choose an appropriate service which suits his/her requirements. The user has to pay according to the amount of resources used in a particular duration of time. This is why it is known as a pay-as-yougo model. It has opened up many new opportunities for researchers as well as business organizations. It is flexible as the users can scale up or scale down resources depending on their requirement.

Various scientific applications face the problem of starvation. Starvation is a condition where a task is in the ready queue, but due to some reason, is not able to get resources. Scheduling algorithms which are priority-based generate such conditions. In such scheduling algorithms, a job with low priority might have to wait for a long time in order to get the resources like processor or input output resources. Efficient scheduling algorithms are required in order to solve this problem thereby reducing the scheduling overhead cost and execution time.

This report discusses various scheduling algorithms and proposes a new scheduling algorithm known as improved minmin scheduling algorithm. The proposed scheduling algorithm considers task size as well as task arrival time. Hence it solves the problem of starvation faced by large sized tasks in the original MinMin scheduling algorithm. The proposed algorithm also performs prediction based on the execution time of the tasks.