基于 DeepSeek 的监控观测平台技术路径探索-智能观测进化论沙龙.pdf

1、基于 DeepSeek 的监控观测平台技术路径探索观测云产品技术总监 黄小龙大模型驱动的监控测性平台技术路径探索1.什么是可观测性和可观测性平台？2.构建大模型驱动的可观测性平台要具备什么条件？3.大模型驱动的可观测性平台有哪些可以落地的场景？CodeTestDeliveryRunOptimizeMonitorOperationUnderstand UserUser SupportUnderstand BusinessDevBusinessOpsUnified Data Collection AgentUnified Data AnalysisPowerful Data Visualisati

2、onMulti-Channel NotificationEnterprise-Grade IntegrationTrueWatch AIHigh Programmbility and ExtensibilityDelivery&Optimisation Backend Performance Monitoring Web App Performance Monitoring Mobile App Performance Monitoring Log Analysis Continuous Profiling Monitoring&Operation Infra Monitoring Docke

3、r Monitoring Network Monitoring APM Synthetic Tests Log Mgmt Metric Mgmt Service Catalog Cloud Cost MgmtUser&Business Analysis User Behaviour Analysis User Session Replay Click Heatmap and Retention Analysis Integration with Business Data,e.g MySQL,OUnified Observability Data Lake 什么是可观测性？“Observabi

4、lity”本身是一种理念，它描述的是通过系统生成的数据来理解系统内部运行状态的能力。DataContext+（数据）（关系）什么是可观测性平台？“Observability Platforms”是指一套工具或系统，它使得-对系统内部状态的外部观察和理解的“Observability”-变成可能。对于可观测性平台，通常包括：Data SourcesData CollectionObservability Data SourcesData ProcessingObservability Data PipelinesData StorageObservability Data Warehouse|D

5、ata LakeO11y ProductsData ExplorersData DashboardsData AI/MLData Monitors基于分散的监控系统落地AI可观测性平台的挑战分散式的监控系统挑战解决方案高质量的数据是构建AI可观测性平台的必备条件数据来源众多，整合难度非常大数据之间缺乏统一的关联标识海量数据存储成本高，难以实时洞察数据查询语义混乱，关联洞察艰难智能 AI可观测性平台统一数据采集统一数据采集统一数据模型统一数据模型统一数据标签统一数据标签统一数据存储统一数据存储统一数据查询统一数据查询AI驱动的监控观测平台必备条件（一）：统一数据采集ClientServiceMi

6、ddlewareInfraThird-partyeventstracelogsmetricsModulesInfraDatabaseMiddlewareCloud NativeLogRUMAPMText line.Observability DataText.Pipeline ScriptOut-of-the-box ScriptsCustomize ScriptsPipeline Build-in Functionsreplace()url_decode()set_tag().Grok()Json()Custom Data Processing Templatesadd_pattern(na