摘要
针对医院信息化系统在容器化部署中“医疗业务中断风险高、数据安全防护弱、能耗管控难”等痛点,本文提出一种融合eBPF内核级采集技术、无监督异常检测与医疗业务适配的多维度容器监控系统。该系统面向HIS(医院信息系统)、LIS(检验信息系统)、PACS(影像归档和通信系统)等医院核心业务容器,实现CPU、内存、网络I/O、医疗数据操作以及能耗的多维度实时监控。通过模拟医院早高峰挂号、集中影像诊断、检验数据传输等场景的干扰实验,分析“容器数量、医疗业务类型、硬件能效”对业务性能以及能耗的干扰规律,建立容器性能-能耗双维度干扰预测模型。该系统具有低侵入性的特点,可适配不同业务类型的应用,为医院后续医疗容器资源管理提供数据支撑。
To address the challenges associated with the containerized deployment of hospital information systems—including a high risk of medical service disruptions,inadequate data security protection,and difficulties in refined energy consumption management,this study proposes a multi-dimensional container monitoring system integrating eBPF kernel-level data collection,unsupervised anomaly detection,and medical business adaptation.The system targets core hospital business containers such as HIS(Hospital Information System),LIS(Laboratory Information System),and PACS(Picture Archiving and Communications System)—to enable realtime,multi-dimensional monitoring of resources.This includes CPU usage,memory usage,network I/O,medical data processing,and power consumption.To simulate a realistic workload of hospitals,interference tests are conducted in the context of e.g.,rush hour at registration desk,bulk image diagnostics sessions,and batch lab data transfers.These experiments analyze the interference laws of container quantity,the nature of medical workload,and the energy-efficiency of underlying hardware on service performance and power consumption.Accordingly,this study constructs a container performance-energy consumption dual-dimensional interference prediction model.The system is non-intrusive and can be adapted to applications of different business types,providing data support for the subsequent management of medical container resources in hospitals.
作者
杨诗音
孙海东
王祥号
王其晓
朱增桂
Yang Shiyin;Sun Haidong;Wang Xianghao;Wang Qixiao;Zhu Zenggui(Lianyungang Hospital of Traditional Chinese Medicine,Lianyungang,Jiangsu 222000,China)
出处
《计算机时代》
2026年第3期65-69,共5页
Computer Era
关键词
医院信息化
eBPF
无监督异常检测
容器监控
性能干扰
Hospital Informatization
extended Berkeley Packet Filter(eBPF)
Unsupervised Anomaly Detection
Container Monitoring
Performance Interference
