The Information Technology (IT) Industry has been revolutionized through Cloud Computing by offering dynamic computing services to a large number of users through its on-demand provisioning of scalable and virtualized resources over the internet on a pay-per-use measured basis. The key features of cloud computing include virtualization, on-demand self-service, elasticity, scalability, and pay-per-use model. The performance of cloud computing services mainly depends on the processing of the jobs submitted to the cloud datacenters by the users. Amount of jobs in the cloud datacenters is very high and the size of jobs can be also massive at any instant. All these jobs need to be executed simultaneously which makes it difficult for the cloud data centers to efficiently handle and manage. Hence, one of the most important research issues which require focus is job scheduling. Scheduling is an integral process in cloud computing whose performance improvements can have a great impact on the efficiency of cloud computing. The scientific research community has been extensively experimenting several scheduling approaches with the aim of optimizing several performance metrics by using several heuristics and meta-heuristic methods. In fact, different scheduling heuristics favour different performance metrics. There has been significant research focusing on the blending of scheduling heuristics and meta-heuristics in different combinations. However, most of the discussed scheduling methods are evaluated on small problem instances and only a few scheduling methods evaluated their scheduling performance on large problem instances. In this Master Thesis, we propose a hybrid job scheduling approach based on a previous work, involving the meta-heuristic optimization technique, genetic algorithm which aims to produce an optimal order based on which several scheduling heuristics should run. This approach is developed with the aim of optimizing the performance metrics namely makespan, flow time, throughput, and total waiting time. Both single-objective and multi-objective optimization based on these metrics has been applied to investigate and identify the best-performing optimization criteria. The developed approach is emulated and evaluated on CloudSimPlus simulation framework using large-scale real-world problem instances (parallel workload benchmarks). The proposed approach is evaluated against several individual scheduling heuristics and the evaluation results suggest that the proposed hybrid scheduling approach outperforms

the individual heuristics in terms of the aforementioned metrics on large-scale problem instances, thus providing better system efficiency and customer satisfaction.