International Journal of Engineering Research and General Science Volume 5, Issue 2, March-April, 2017 ISSN 2091-2730 50 www.ijergs.org Review on Meta-heuristic Scheduling Optimization Techniques in Heterogeneous Clouds Shubhdeep Kaur Sandhu, Anil Kumar (Asst. Professor) CET Department, GNDU, Amritsar, sandhu.shubh92@gmail.com Abstract— As cloud computing is turning out to be evident that the eventual fate of the cloud industry relies on interconnected cloud systems where the resources are probably going to be provided by various cloud service suppliers. Clouds are also seen as being multifaceted; if the user requires only computing capacity and wishes to personalize it as per his requirements, the infrastructure cloud suppliers are able to provide this convenience as virtual machines. Many optimized meta-heuristic scheduling techniques are introduced for scheduling of bag-of-tasks applications in heterogeneous framework of clouds. The overall analysis demonstrates that, utilizing different meta-heuristic techniques can offer noteworthy benefits in the terms of speed and performance. Keywords— Bag-of-tasks, Heterogeneous clouds, Meta-scheduling, Meta-heuristics, Simulated Annealing, Tabu Search, Multi- criteria Decision Making 1. INTRODUCTION 1.1 Cloud Computing Within the course of the past couple of years, cloud computing has come forth as a standout amongst other solutions for delivering IT oriented services to the clients. It is the novel concept with the help of which services are distributed amongst consumers and providers after identifying the customer demands and sandboxing their requirement in virtualized settings [12]. From the infrastructure point of view, Cloud Computing is propitious resolution that extends the resource capacity of independent computing systems dynamically. Cloud computing is analogous to Grid computing in the manner that it also deploys the distributed resources to attain application-level targets [8]. Its proficiency to leverage virtual technologies at the hardware level as well as application level in order to recognize the properties of sharing the resources, providing dynamic resource scaling “o n-demand” while offering a flexible price framework in conjunction with ease of modification and high availability makes it superior to the Grids. On the other hand, with the help of utility based price frameworks and on-demand resource as well as service provisioning, service suppliers can maximize the resource utilization along with minimization of operational cost. A service provider does not need to offer capacities in accordance with the peak load anymore, which results in magnificent savings when the resources are set free to save operational costs in case service request is reduced [8]. 1.2 Inter-cloud Systems The term “inter-cloud” implies an interoperable environment in which multiple criteria collude to satisf y QoS levels [12]. Once the multiple clouds are interlinked together, different clouds provide dissimilar architectures and varying resources which are consolidated into a single entity in a transparent manner [16]. Inter-cloud intends to expand the service elasticity of cloud and scalability while minimizing the performance and service cost overheads [15]. Inter-cloud systems support dynamic workload supervision to initiate decision making for job distribution at meta-brokering level. Inter-cloud meta-broker is built to be decentralized and dynamic by improving the way choices are made for service distribution [12]. This can be carried out through the use of heuristic criterion and algorithms to achieve improved meta-scheduling in inter-cloud environments. In each scheduling decision, percentage of required resources is ought to be reconfigured, displacing them to an alternate cloud region. This course of action causes some virtual machines to be paused for a short time period, which in turn can cause performance degradation temporarily [10]. 1.3 Meta-scheduling Paradigms We pay attention to performance optimization using meta-scheduling paradigm to attain a much better job scheduling across multiple clouds. When numerous distinct clouds are merged, a multi-layered technique is needed that ought to have a universal scheduler, which manages the allocation of jobs amongst the clouds in addition to the ones that are local cloud schedulers [16]. The meta-broker