University of Khartoum

Implementation of an Earliest Deadline First Algorithm in Cloud Computing Environment using The CloudSim Tool

Implementation of an Earliest Deadline First Algorithm in Cloud Computing Environment using The CloudSim Tool

Show full item record

Title: Implementation of an Earliest Deadline First Algorithm in Cloud Computing Environment using The CloudSim Tool
Author: Ali, Sayda Khidr Fadlalah
Abstract: This project implements an earliest deadline first (EDF) algorithm which is a dynamic real time scheduling algorithm in cloud computing, with the objectives of design a cloud broker using CloudSim tool to schedules cloudlets (tasks) that have time constraints to hosts by ordering each of which by deadline that is estimated proportionally to the cloudlets length, and examine the cloudlets performance assigned to different number of hosts, and test the deadline value variation affect on the cloudlets performance. After that analyze the cloudlets (tasks) performance results under this broker and compare it with first come first served Scheduling policy to evaluate the performance. In CloudSim simulated data center the number of resources with their characteristics remained fixed except for the number of hosts and the deadline value which increased several times, three cases were taken 100 hosts, 150 hosts and 200 hosts and every time processed 1000 cloudlets assigned by the designed EDF broker and the FCFS policy, the performance results of the cloudlets (task) at each case are analyzed, also the deadline is tested in three values 1.15, 1.25 and 1.5. The results analysis showed 19.4% cloudlets missed the deadline under 100 hosts, 44% under 150 hosts and zero under 200 hosts when assigned by EDF broker against 30.7%, 52.7% and 12.2% when assigned by FCFS policy using 1.25 as deadline, also this results reduced the waiting time and the wall time and the delay after deadline more for each cloudlets that assigned by the EDF broker. the results of the deadline values showed that most cloudlets does not have much time to finish before the deadline 1.15 but it could affect the performance greatly with a bigger number of hosts although 1.5 deadline value showed a better results than the two past values in which in 150 and 200 number of hosts cases no cloudlets missed the deadline, also the number of hosts, and virtual machines needed for any cloud infrastructure could be determined by this scheme in order for all cloudlets to meet the deadline while the wall time and the waiting time remain acceptable
Description: 94page
URI: http://khartoumspace.uofk.edu/123456789/16943
Date: 2015-11-10


Files in this item

Files Size Format View

This item appears in the following Collection(s)

Show full item record

Share

Search DSpace


Browse

My Account