Research On Privacy And Security-aware Workflow Scheduling In A Hybrid Cloud

Cloud computing is well-known for its characteristics of ‘high elasticity and pay-as-you-go’.Because it can provide nearly "infinite" service instances and a high-performance heterogeneous distributed environment,increasing numbers of workflow applications are deployed onto the cloud environment for execution.As an emerging cloud environment,hybrid cloud is a combination of private cloud and public cloud,which are connected through the Internet.When deploying workflows in a hybrid cloud environment,two concerns to be considered urgently:(1)privacy-sensitive data and tasks cannot be exposed on public cloud platforms;(2)data transmission security across public and private cloud platforms via the Internet should be guaranteed.The workflow scheduling in the existing hybrid cloud environment usually only focuses on one problem,or only regards the hybrid cloud as an extension of the cloud environment.Therefore,this thesis studies how to effectively reduce workflow execution costs in a hybrid cloud environment under the constraints of deadlines,while meeting the privacy and security requirements of workflows.Concerning the inadequacies of study on privacy and security in the hybrid cloud environment in the existing research,this thesis uses hybrid encryption and hash functions to ensure the security and integrity of data transmission in the hybrid cloud environment,respectively.Then a three-level data security model is established to ensure the secure transmission of data at different levels in the hybrid cloud environment.Based on the three-level data security model,tasks are divided into two categories: tasks with high privacy and security requirements,namely private tasks;tasks with low privacy and security requirements,namely public tasks.Based on the data security model and task division rules,the constrained optimization problem in this thesis is modelled.Based on the above model,we first design a privacy and security-aware list scheduling algorithm.It assigns a user-given deadline to each task in the workflow,and then selects a favorable service instance that satisfies privacy constraints and minimizes execution cost for each task in the workflow,and finally obtain a scheduling plan with a lower cost on the premise of not violating the deadline.Based on the proposed list scheduling algorithm,we use the simulated annealing algorithm to further improve the effect of the scheduling scheme,and obtain a scheduling scheme with lower execution cost by changing the task sequence in the iterative process.We conduct simulation experiments on real-world workflow datasets and results show that the two proposed algorithms outperform the existing four algorithms.The algorithm of simulated annealing combined with list scheduling outperforms other algorithms at the expense of a certain amount of execution time,the cost is reduced by 12.53% compared with the list scheduling algorithm in this thesis.