Research On Scientific Workflow Scheduling Algorithm Based On Cloud Computing

As a service,cloud computing effectively solves people’s needs for the Internet,and has been widely used in education,medical care,transportation,government affairs,media,logistics and other industries.But cloud computing also faces many challenges due to the uncertainty and diversity of user-submitted tasks,as well as the complexity and variability of the network environment.At present,in order to improve the performance of cloud computing,many scholars focus on the optimization of task scheduling under cloud computing.Because most of the algorithms used to deal with task scheduling problems are heuristic algorithms,but this type of algorithm has the disadvantages of poor stability,optimization performance depends on the problem and experience,so it is of certain practical significance to study an efficient task scheduling algorithm.Focusing on the task scheduling problem of cloud computing,an improved particle swarm algorithm MSMPSO is proposed.The algorithm mainly includes three types of strategies: due to the pseudo-random characteristics of the random number algorithm of modern CPU,the initialization of particle swarm is uneven.To solve this problem,Sobol sequence is used to initialize the position of particle swarm.Aiming at the problem that the parameter inertia weight of the standard particle swarm algorithm cannot effectively balance the global search and local search capabilities of the particle swarm,the inertia weight is dynamically adjusted by the method of mixing cosine and exponential functions.In addition,in order to further balance the search ability within the particle swarm,the particles are sorted and classified according to their fitness,and inertia weights controlled by different acceleration factors are set for the particles of different levels.Aiming at the shortcomings of particle swarms that are prone to premature convergence and difficult to jump out of local optimality,the concept of local optimality is defined by combining variance and the invariant times of optimal population solutions,and the Cauchy function is used to mutate optimal population solutions to help particles The group jumps out of the local optimum.On the WorflowSim platform,five types of workflows,including computing intensive and CPU intensive,were tested using the completion time and execution cost(transmission cost and compute cost)of the scientific workflow as indicators,and compared with four algorithms such as the standard particle swarm algorithm.Through the analysis of experimental results,it is found that the completion time of the MSMPSO algorithm can be reduced by up to 51% and the cost by up to 50% compared with several other algorithms,which verifies the efficiency of the MSMPSO algorithm in handling the workflow scheduling problem.