Cloud computing as a pattern for distributed computing, are composed of large shrimp ask combined resources with the goal of resource sharing as a service, on the internet. Such resources as in memory, processor and services are always worth and more efficient use of these, is endless challenge
Hence the scheduling of tasks in cloud computing is very important that try to determine an efficient scheduling and source allocation. In fact, the goal is determining a processing resource from set of resources that a task needs for process, so that can process more jobs in less time. Scheduling system controls different functions in cloud system for increasing job completion rate, resource efficiency and in consequence increasing the computing power. In this study, we provide approach base on lottery algorithm to reach those goals with minimize make-span time. Simulation of proposed method is by “CloudSim” application
The Proposed method:
As noted earlier in task timing issues in cloud computing, the output is a proper mapping of tasks to resources. So that parameters such as response time, make-span time and performance of data centers, are optimized. In this report, we present a new algorithm based on lottery algorithm. The proposed algorithm has evolutionary view and most prominent characteristic of it is being agile. Stages of it are as follows:
First step: number of answers is created randomly, which either is discussed as a participant in proposed method.
Second step: the propriety of each participant is measured :
Fi=value obtained for task i Tlen=length of ith task
Rw=workloads that are already on the CPU which ith task is allocated for it
NCC=communication cost for the selected virtual machine for ith task or data broker Fitness total=the answer fitness
Third step: each participant assigned a fitness based on the number of sheets. Indeed, participants who had more points will have more sheets. Forth step: lottery is done. There is one win rate equal to 0.8. indeed, what it means is that in each iteration 80% of current stage’s participants are transported to new stage. 20% of the initial population consist of participants. For lottery, we have used randomly numbers with uniform dispatch. Fifth step: we check end condition in this stage. In our proposed method, the end condition is specific number of repetition, but we can consider the end condition near an optimal condition.