Eureka! Institutional Repository

Cloud Computing is a new technology in the academic world. On Cloud Computing platform of, resources are provided as as services , providing the agreed services by the Service Level Agreements (SLA), agreed with subscribers. However, due to the fact that these services are shared heterogeneously , and due to the dynamic heterogeneity among individuals and platforms, resource waste is inevitable. Task scheduling and load balancing, is part of the problem in the area of Cloud Computing, which records the assignment of tasks to the Virtual Machines with the best possible way. In this thesis we introduce a task scheduling evolutionary algorithm and compare it, with a First Come First Served linear algorithm, the most common algorithm for task scheduling on cloud computing. The purpose of applying an evolutionary algorithm, is to record the total execution time of all tasks assigned to the virtual machines for different sets of tasks. Different configurations of the evolutionary algorithm are compared. The performance of the evolutionary algorithm is also compared with the FCFS algorithm. To solve the problem of outsourcing we recommend applying the evolutionary algorithm by calculating the fitness value using standard deviation. All the results and the methods that were created are presented in details. In the end, the results of the two algorithms are compared using Cloudsim, a Cloud Computing simulation program.