人工智能代理和传统工作负载的可观测性.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.C O P 3 3 5Observability for AI agents and traditional workloadsIgor SedukhinHe/HimGM Application ObservabilityAWSSiva GuruvareddiarHe/HimSenior Specialist Architect

2、 AWSMuthukumaran ArdhanaryHe/HimSenior Director SRECCC Intelligent Solutions 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Common Challenges-ApplicationComplex Microservice ArchitecturesLack standard application metricsDifficult to prioritize with business goalsDisjointed monito

3、ring experiencesApplications are complex and need to be continuously available and performant to maintain end-user satisfactionManually stitch together telemetry signals to determine which services and APIs have degraded availability or high latenciesHard to know which anomalies matter most to the b

4、usiness and require immediate attentionCorrelating telemetry signals from disjointed experiences across metrics,traces,logs,real-user monitoring,and synthetic monitoring lead to extended time to detect and recover 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.CloudWatch Applicat

5、ion SignalsService Level ObjectivesSet reliability goals based on business objectivesStandard application metrics golden signalsPre-built,standardized dashboard with standard application metricsMetrics to root cause in 3 clicksRate of failed requests due to malformed request data.Discover Services a

6、nd TopologyWith no manual effort and no custom code,leveraging OpenTelemetry SDKs 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.HOW IT WORKSAPPLICATION SIGNALSJavaPythonAWS Distro for OpenTelemetry SDKsDiscover and visualize application topologyPre-built dashboards to assess hea