Arrival Based Deadline Aware Job Scheduling in Cloud

Abstract

Cloud Computing has been gaining much attention in recent years due to the fact that many real world applications have become more complex and dynamic. Cloud Computing provides its users with facilities where a user need not to pay for the whole license of a software or to buy a huge storage, a user can only pay according to his demand. Cloud Computing is the delivery of services over the Internet according to user's demands. It can basically be de ned as anytime, anywhere, through any device accessing of various services through the Internet. Due to advances in Cloud Com- puting, the user base of cloud computing is increasing hugely and therefore there is need of scheduling algorithms in order to schedule all those resources provided by cloud server. This study contains a brief review of a scheduling algorithms for the resources in cloud environment and discuss their pros and cons. To make the algorithms more applicable to real time scenario in order to schedule the resources according to user de ned parameters and to make it more bene cial, a new variant of these algorithm is proposed which aims to assign jobs in clouds in order to nish more and more jobs within the deadline constraint provided by its users. Scheduling of resources according to deadline is an important measure as there may be jobs which needs to be done before time and have no e ect after the deadline is over so an algorithm which resolve this need is to be developed in order to make cloud computing services more e ective. The objective is to propose the algorithm for deadline constrained jobs and also de ne its bene ts over other algorithms. Further, it de nes the di erent param- eters that can be used to compare an algorithm and uses those parameters to compare the algorithms. The algorithm developed here is named as Arrival based Deadline First Job Scheduler(ADSF) and is compared to First Come First Serve Algorithm(FCFS) in terms of waiting time, the delayed number of jobs and the total delay time. The results show a signi cant improvement in the former as compared to latter.