محمدرضا استادي مقدم

With the expansion of the Internet and the increasing need of users for computing resources, new models of using hardware and software resources have been noticed in recent years. In the meantime, cloud computing has been welcomed as a new model of providing, receiving and delivering services due to its high potential. One of the upcoming challenges in the cloud computing industry is providing resources and scheduling tasks. Task scheduling plays a key role and affects the efficiency of the cloud environment and the quality of service provided to users. Considering the expansion of public interest in cloud computing in recent years, and on the other hand, the increase in the graph of tasks entering the cloud environment; Other classical methods are not able to find an optimal schedule. Therefore, in this thesis, we are trying to design and introduce a suitable scheduler in order to schedule tasks and optimal management of resources by using the knowledge of reinforcement learning and neural networks, in addition to adapting to the needs current schedules; lead to a reduction in the cost of electricity consumption of resources. In order to implement and eva‎luate the proposed method; We have used Google Cluster data. Therefore, in the real world, there are dependencies between tasks; With the help of random data, we have randomly created dependencies between the tasks of this data set. The number of input tasks between 1000 and 5000 and the number of servers between 100 and 300 have been simulated in different scenarios. The results of the implementation show that the proposed method has succeeded in reducing the cost of energy consumption by 69.20% compared to the classical Round-Robin method. Also, compared to the other smart method discussed in this thesis, the proposed method has succeeded in reducing the cost of energy consumption by 28.74%.

زمان بندي وظايف , رايانش ابري

Task Scheduling , cloud computing

Mohammad Reza Ostadi Moghaddam

Dr.Hadi Shahriyar Shah Hosseini