《计算机应用研究》|Application Research of Computers

基于特征选择的VNF资源需求预测方法

VNF resource demand forecast method based on feature selection

免费全文下载 (已被下载 次)  
获取PDF全文
作者 武静雯,江凌云,刘祥军
机构 南京邮电大学 通信与信息工程学院,南京 210003
统计 摘要被查看 次,已被下载
文章编号 1001-3695(2021)10-044-3131-06
DOI 10.19734/j.issn.1001-3695.2020.12.0561
摘要 针对在网络切片场景下以往的VNF(虚拟网络功能)资源分配策略无法满足动态的资源需求,很容易导致资源分配不足或过度分配的问题,提出了一种基于两阶段算法(two-stage algorithm,TSA)的VNF资源需求预测方法。该方法首先基于数据特征筛选出与预测目标高度相关的候选特征集,然后利用贪婪式前向搜索策略对候选特征集进一步筛选获得最优特征集,最终训练出不同类型的预测模型。仿真结果表明,基于该方法所训练的模型可以获得更好的预测性能,同时该方法的可扩展性较好,训练好的模型可以直接集成到现有的VNF部署算法中应用。
关键词 网络切片; 网络功能虚拟化; VNF资源分配; 特征选择
基金项目 国家自然科学基金资助项目(61871446)
江苏省重点研发项目(BE2020084-4)
本文URL http://www.arocmag.com/article/01-2021-10-044.html
英文标题 VNF resource demand forecast method based on feature selection
作者英文名 Wu Jingwen, Jiang Lingyun, Liu Xiangjun
机构英文名 College of Telecommunications & Information Engineering,Nanjing University of Posts & Telecommunications,Nanjing 210003,China
英文摘要 In the network slicing scenario, the previous VNF(virtual network function) resource allocation strategy can't meet the dynamic resource demand, which can easily lead to the problem of insufficient or over-allocation of resources. To solve this problem, this paper proposed a VNF resource demand prediction method based on two-stage algorithm(TSA). The algorithm first filtered out candidate feature sets highly related to the predicted target based on data features, and then used a greedy forward search strategy to further filter candidate feature sets to obtain the optimal feature set. Finally, it trained the different types of prediction models through different regression algorithms. The simulation results show that the model trained by the proposed algorithm can obtain better prediction performance, and the algorithm has better scalability. The trained model can be directly integrated into the current VNF deployment algorithms for application.
英文关键词 network slicing; network function virtualization; VNF resource allocation; feature selection
参考文献 查看稿件参考文献
 
收稿日期 2020/12/23
修回日期 2021/2/4
页码 3131-3136,3142
中图分类号 TN929.5
文献标志码 A