Deadline Constraint-based Workflow Task Scheduling In Cloud Environment

The dynamic,heterogeneous and unpredictable nature of cloud computing,makes the resource allocation and task scheduling strategy of cloud facing great challenges.In addition,with a large number of users asking for the quality of service(QoS),it is necessarily to develop a proper and effective scheduling scheme,in order to make the most use of cloud computing resources.The PSO(Partical Swarm Optimization)is one of the most popular algorithm applied to this area this years.But the traditional PSO has some disadvantages such as slow convergence and easy to fall into local optimization.Besides,existing scheduling research has some shortcomings:They are only focus on minimizing either execution time or cost considerations,ignoring the heterogeneity and performance fluctuations of resources,or ignoring the data transfer times.This paper applied the workflow management technique to the cloud computing task scheduling,while considering users' deadline demands.We mainly studied the contents as follows:Firstly,Considering the shortcomings of the traditional PSO,we proposed an improved particle swarm optimization with dual fitness(DFPSO).We considered the cloud computing workflow execution time and total cost as the time fitness function and cost fitness function respectively.Besides,we designed the historical and global optimization rules on dual fitness,according to the scheduling purpose we are focusing on.Secondly,By considering the workflow execution time control in cloud environment,we designed a workflow execution algorithm.Combining DFPSO algorithm,we finally designed a deadline constraint-based workflow task scheduling strategy in cloud environment.Our strategy considered the heterogeneity and performance fluctuation of the virtual machines,as well as the transmission times between tasks.The scheduling objective is to meet the deadline constraint while minimizing the total cost.As for the experiment,we used CloudSim as the simulation platform,and selected four typical scientific workflow as the evaluation workflow.We used ICPCP and PSO as the comparing algorithm.We analyzed the performance in execution time and total cost of the three scheduling methods in the condition of different deadline-constraints.The results showed that the DFPSO had a better optimization performance.It could better meet the deadline,and save lots of costs while reducing execution time.