摘要
铁路综合视频监控系统承担着大量高清视频的分析处理及检索查询,对计算资源的平滑扩展和负载均衡提出了很高要求。作为容器云的主流编排平台,K8s能很好地满足铁路综合监控系统的绝大多数应用需求,但在资源的平滑扩展和均衡性上存在不足,需要有针对地进行优化。通过详细分析综合视频监控系统的具体需求以及K8s固有调度机制的不足,采用自研的应用管理程序和K8s就地垂直伸缩机制相结合的优化方案,不仅能够快速且准确地判断Pod资源动态平滑扩展时机,自动进行Pod资源动态平滑扩展,而且还能自动使集群中各个工作节点资源占用更加合理均衡,从而有效减少人工维护工作量,降低操作风险,提高整个系统的可靠性和可用性。
The railway integrated video surveillance system undertakes the analysis,processing,retrieval,and inquiry of a large number of high-definition videos,which puts forward high requirements for the smooth expansion and load balancing of computing resources.As the mainstream orchestration platform for container clouds,K8s can effectively meet the vast majority of application requirements for railway integrated surveillance systems.However,it has deficiencies in smooth expansion and balance of resources,which require targeted optimizations.Through a detailed analysis of the specific requirements of the integrated video surveillance system and the deficiencies of the K8s inherent scheduling mechanism,an optimization solution combining self-developed application management program and K8s in-place vertical scaling mechanism is adopted.It can not only quickly and accurately determine the timing for dynamic smooth expansion of Pod resources,automatically perform dynamic smooth expansion of Pod resources,but also automatically enhance the reasonability and balance of resource occupation of each work node in the cluster,thus effectively reducing manual maintenance workloads,minimizing operational risks,and improving the reliability and availability of the entire system.
作者
李永康
林海
何伟
董成文
LI Yongkang;LIN Hai;HE Wei;DONG Chengwen(Standards and Metrology Research Institute,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China;Henan Splendor Science&Technology Co.,Ltd.,Zhengzhou Henan 450000,China)
出处
《中国铁路》
北大核心
2025年第8期93-98,共6页
China Railway
基金
中国铁道科学研究院集团有限公司科研开发基金项目(2024YJ290)。
作者简介
第一作者:李永康(1987-),男,副研究员。Email:690834588@qq.com;通信作者:林海(1978-),男,工程师。Email:linhai@hhkj.cn。
