Skip to content

AIP-C01 · Study Guide

← Domain 4: Operational Efficiency and Optimization for GenAI Applications

4.2.6 · Lesson 10 of 16

Profile Bottlenecks Across FM, Retrieval, and Tools

Find where latency and cost accumulate across orchestration steps using distributed tracing and log analytics.

WHAT

Profile bottlenecks across FM calls, retrieval, and tools in GenAI workflows.

HOW

  • AWS X-Ray: Trace API Gateway → Lambda → Bedrock → external tools.
  • CloudWatch Logs Insights: Analyze prompt-to-token ratios, token spikes, and max-token truncations.
Client → API Gateway → Lambda → Bedrock → Tools
                 ↑
              X-Ray trace

Key points

  • AWS X-Ray across API → Lambda → Bedrock → tools
  • CloudWatch Logs Insights
  • Prompt-to-token ratios
  • Max token truncation diagnosis
  • Multi-agent bottleneck finding

Exam focus

X-Ray answers where the request was slow across services; Logs Insights analyzes token spikes and truncation in invocation logs.

View progress