摘要
针对大数据引擎完成计算功能时易出现负载不均衡问题,提出一种大数据计算引擎均衡部署数学建模方法。结合流式计算与批量计算模式,构建双模式大数据计算架构。以公平为前提合理分配引擎任务数量,获取剩余计算容量,设定带权思想负载均衡约束,根据数据流大小,确定部署最佳开销;以约束模型为部署目标,利用蚁群算法初始化处理信息素函数,得到引擎被部署到各任务节点的几率,通过启发函数确定蚂蚁爬行路径,经过信息素的局部和全局更新,完成所有路径循环,得出全局最优解,计算引擎最佳部署方案。仿真结果证明,上述方法所用部署时间最短,可实现计算引擎的均衡负载,减少部署开销。
Aiming at the load imbalance of big data computing engines,this paper puts forward a mathematical modeling method of balanced deployment for big data computing engines.Firstly,a dual-mode big data computing architecture was constructed by stream computing and batch computing.On the premise of fairness,engine tasks were allocated reasonably to obtain the remaining computation capacity,and then the weighted load balancing constraint was set.Moreover,the optimal deployment cost was determined by the size of the data flow.After taking the constraint model as the deployment target,the ant colony algorithm was used to initialize pheromone functions,and thus to get the probability that the engine was deployed to each task node.Furthermore,the moving paths of ants were determined by the heuristic function.Through local and global pheromone updates,the loop of all path circulation was completed,and the global optimal solution was obtained.Finally,the best scheme of engine deployment was designed.Simulation results show that the proposed method has the shortest deployment time,achieves the load balance of the computing engine and reduces the deployment cost.
作者
孟小燕
赵希武
MENG Xiao-yan;ZHAO Xi-wu(Inner Mongolia Normal University,Hohhot Inner Mongolia 010051,China;School of Computer and Information Engineering,Inner Mongolia Normal University,Hohhot Inner Mongolia 010051,China)
出处
《计算机仿真》
北大核心
2022年第11期472-476,共5页
Computer Simulation
关键词
大数据
计算引擎
均衡部署
数学建模
蚁群算法
约束条件
Big data
Computing engine
Balanced deployment
Mathematical modeling
Ant colony algorithm
Constraint condition