论文部分内容阅读
资源管理系统就是对计算机集群网络中分散在各个节点中的多种资源统筹管理,资源需求预测和任务调度是其核心技术。提出了一种基于用户资源需求预测的相似度匹配(SDM:Similarity Degree Matching)算法,首先对用户将来发送任务的各种资源需求量进行预测,再采用基于任务与节点相似度匹配的调度算法,综合任务资源需求与节点服务能力两方面而得出相似度匹配,从而实现任务到节点的映射。C++语言环境下仿真结果表明,该算法实现了服务节点内部的各种资源负载平衡,提高系统的吞吐量和使用率。
Resource management system is a cluster of computer network nodes scattered in a variety of resources in a unified management, resource demand forecasting and task scheduling is its core technology. This paper proposes a Similarity Degree Matching (SDM) algorithm based on user resource demand forecasting. Firstly, it predicts the amount of resources required by the user to send tasks in the future, and then uses a scheduling algorithm based on matching of task and node similarity. The similarity matching is obtained by combining the task resource requirements with the node service capabilities, so as to realize the task-to-node mapping. The simulation results in C ++ language environment show that this algorithm realizes the load balancing of various resources inside the service node and improves the system throughput and usage rate.