A Study Of Energy- And Reliability-aware Cost Optimization For Cloud Workflows

The execution of complex scientific workflow applications on cloud typically involves a large number of virtual machines(VMs),which makes the cost as well as the energy consumption a great concern.To partially alleviate this issue,some cloud service providers such as Cloud Sigma and Elastic Hosts introduce new pricing policies where users are charged on the basis of allocated CPU frequencies together with various combinations of VM configurations and prices.However,the customizable CPU frequencies make resource provisioning and scheduling harder to achieve a cost-optimal VM scheduling solution.As the higher CPU frequencies cause high-energy consumption with enhanced reliability,while lowering the CPU frequencies,to reduce the energy consumption,produce soft error issues,which results in a high rate of completion time failures of workflow applications.Therefore,a frequency tuning approach is highly desirable to obtain a cost-optimized workflow scheduling solutions.To address the above challenges,this study proposes a novel method based on genetic algorithms.By adopting our newly introduced genetic operators(i.e.,crossover and mutation),our approach can quickly figure out a cost-optimal resource provisioning and task scheduling solutions for cloud workflows by allocating tasks to appropriate VMs with specific operating frequencies under the energy,reliability,makespan and memory constraints.This thesis makes three main contributions as follows:1.We formalize the cost-optimization problem of task scheduling for cloud workflows under the constraints of makespan,energy,memory,and reliability while considering the overhead of checkpointing with rollback-recovery.2.Based on the genetic algorithm,we introduce a new genetic operator namely chromosome modification that satisfies the memory constraint,where the memory required for a task should not exceed the limit of VM's RAM on which tasks are executing.3.We propose an effective approach by employing a new set of techniques that can quickly find out a near cost-optimal resource provisioning and task scheduling solutions for workflow applications by allocating tasks to suitable VMs.Moreover,we introduce a frequency tuning scheme to achieve a cost-optimal solution while satisfying other Qo S requirements.Extensive experiments on various well-known scientific workflow benchmarks validate the effectiveness of the proposed method.Comparing with state-of-the-art methods,our approach can not only reduce the overall cost but also can achieve energy saving without violating those constraints.