الفهرس 
Cloud Computing BackgroundOnly 14 pages are availabe for public view
 |  | from | 128 |  |  |
| | | from | 128 | | |
Abstract
Cloud computing offers shared processing resources and data by a host application service provider so that the user does not need to buy a server. Scheduling the tasks is essential technique in the cloud environment. It is required to allocate the tasks into the proper resources that optimizes the overall system performance. Particle Swarm Optimization (PSO) algorithm is the most popular scheduling algorithm that is used to maximize resource utilization. However, the performance of the PSO scheduling algorithm decreases in the case of scheduling a large number of tasks. In this thesis, Improved Particle Swarm optimization (IPSO) algorithm is proposed to allocate a large number of tasks to available resources optimally. This can be achieved by splitting the submitted tasks into some batches in a dynamic way to maximize the use of resources. We have used a new parameter µ as the leading indicator for the resources utilization state. Parameter µ represents the mean of available resources in the system. In each creation for batch, the parameter µ is calculated as the primary reference for the utilization check. The proposed algorithm IPSO dynamically splits the tasks into batches with an acceptable consumption of available resources in order to obtain a fast load balancing. After getting sub optimal solution for each batch, IPSO appends all sub optimal solutions for batches into the final allocation map. Finally, IPSO tries to balance the loads over the final allocation map. Implementation of the IPSO algorithm was done using CloudSim simulation tool which is used to model and simulate task scheduling in the large scale cloud computing. The virtual nodes and computing resources were modeled to evaluate the efficiency of the IPSO. The proposed IPSO algorithm is compared with different scheduling algorithms; PSO, Honey Bee (HB), Ant Colony Optimization (ACO), and Round Robin (RR) algorithms. The results of experiments show the efficiency of the proposed algorithm in terms, makespan, standard deviation of load, and degree of imbalance. IPSO minimizes the makespan up to 50% when it works with large scale data. This is due to that IPSO improves the PSO by decreasing the workload on each particle in PSO by batching the big list of tasks into small sub list.