Optimization Techniques For Resource Allocation And Scheduling Of Scientific Workflows In Cloud Computing Environments

Scientific workflow applications are collections of several structured activities and finegrained computational tasks.With the emergence of e-Science,workflow scheduling has become a significant component for the representation of complex multiple dependent tasks and the flow of control between them.Scientific workflows,like other applications,benefit from the Infrastructure as a Service(IaaS)clouds,which offers access to a scalable amount of resources provisioned elastically on demand and are charged on a pay-per-use basis.However,along with these benefits come numerous challenges that need to be addressed in order to perform scheduling of big data applications on cloud resources in an efficient manner.The complexity of intricately connected tasks in scientific workflow applications is compelling researchers to explore heuristic,metaheuristic,and hybrid techniques to get optimum solutions of the workflow scheduling problems.Inefficient resource allocation and scheduling directly leads to higher time and cost for no additional benefit.This thesis proposes novel optimization techniques for resource allocation and scheduling of scientific workflows in cloud computing environments.The methods help to optimize a set of Quality of Service(QoS)requirements such as execution time,execution cost,load balancing,and resource utilization,for both users and cloud providers.This key contributions of this thesis are:(i)identification and description of the challenges particular to cloud environments that scheduling algorithms must address;(ii)a comprehensive taxonomy and survey of state of the art techniques for scheduling scientific workflows in cloud computing environments;(iii)proposed a technique that combines heuristic and Mixed Integer Programming(MIP)model for scheduling of tasks on dynamic and scalable set of Virtual Machines(VMs)in order to optimize cost while satisfying their deadlines;(iv)proposed a novel hybrid metaheuristic for multiobjective optimization of conflicting objectives based on Pareto optimal non-dominated solution to achieve optimum convergence and diversity of the Pareto front;(v)proposed an efficient structure-aware and budget-aware workflow scheduling strategy on dynamically provisioned cloud resources with the objective to optimize the execution time.The techniques presented in this thesis will benefit both cloud providers and users by providing significant improvement in the resource allocation and scheduling of scientific workflows on the cloud.