Help Center/ Well-Architected Framework/ Well-Architected Framework and Practices/ Performance Efficiency Pillar/ PERF04 Performance Analysis/ Collecting Performance Data/ PERF04-05 Collecting Application Performance Data
Updated on 2025-05-22 GMT+08:00
View PDF
Share

PERF04-05 Collecting Application Performance Data

  • Risk level

    Medium

  • Key strategies

    Application performance data (throughput, latency, and completion time) is usually collected by embedding code snippets or integrating tools into application code. The collected application performance metrics enable bottleneck identification, system behavior evaluation, availability risk detection, and capacity planning.

    Common monitoring strategies include:

    • APM tools: Use cloud-based or open-source APM to analyze performance data (metrics, logs, and traces).
    • Log-based tracing framework: Implement frameworks with log generation, formatting, and contextual correlation capabilities. By integrating these into code repositories, runtime performance data can be collected.
    • Customized detection: Develop custom code to collect unique performance metrics only when platform metrics are insufficient.
    • Use observability standards in the industry. Adopt industry standards–based tools like OpenTelemetry.

    Suggestion: Use the distributed tracing technology to identify request links between multiple services and components. Collect tracing data to analyze data flows from end to end, and identify bottlenecks or inefficient request segments for further optimization.

  • Related cloud services and tools
    • Application Operations Management (AOM)
    • Application Performance Management (APM)
    • Log Tank Service (LTS)