Abstract: It is a challenge to reduce data transmission time while ensuring data security for large datasets in scientific workflows.Combining the advantages of cloud computing and edge computing,a replica-based data placement model in edge-cloud environments was developed,and a Nonlinear inertial weight discrete Particle Swarm Optimization algorithm based on Genetic Algorithm operators（NPSO-GA）was proposed.This algorithm considered the factors such as data security,transmission bandwidth and storage capacity of datacenters,and the data replicas to reduce the transmission latency in scientific workflows was generated dynamically.The crossover and mutation operators of the genetic algorithm were introduced into the PSO algorithm to enhance its search ability and prevent prematurely convergence.Experimental results showed the data placement strategy with replica mechanisms based on NPSO-GA could effectively reduce transmission time during the execution of scientific workflows.It is a challenge to reduce data transmission time while ensuring data security for large datasets in scientific workflows.Combining the advantages of cloud computing and edge computing,a replica-based data placement model in edge-cloud environments was developed,and a Nonlinear inertial weight discrete Particle Swarm Optimization algorithm based on Genetic Algorithm operators（NPSO-GA）was proposed.This algorithm considered the factors such as data security,transmission bandwidth and storage capacity of datacenters,and the data replicas to reduce the transmission latency in scientific workflows was generated dynamically.The crossover and mutation operators of the genetic algorithm were introduced into the PSO algorithm to enhance its search ability and prevent prematurely convergence.Experimental results showed the data placement strategy with replica mechanisms based on NPSO-GA could effectively reduce transmission time during the execution of scientific workflows.

Key words: cloud computing, edge computing, scientific workflow, data replica, data placement

摘要： 面对科学工作流中的大型数据集,如何在保证数据安全的前提下最小化传输时延是一个重要挑战。将云计算和边缘计算的优势相结合,构建了云边环境下基于副本的数据布局模型,并提出了一种基于遗传算法算子的非线性惯性权重离散粒子群优化算法（NPSO-GA）。该方法考虑了数据安全、数据中心的传输带宽、存储容量等因素,动态生成数据副本以降低科学工作流传输时延;通过在粒子群算法中引入遗传算法的交叉和变异算子,增强了粒子群算法的搜索能力,避免早熟收敛。实验结果表明,基于NPSO-GA的数据副本布局策略可以有效降低科学工作流执行过程中的传输时延。

关键词: 云计算, 边缘计算, 科学工作流, 数据副本, 数据布局