4.3.2 Profile Application Performance

Profile Application Performance

Amazon CodeGuru Profiler

What it does:

• Profile application performance
• Provide different visualizations of profiling data
• Identify what code is running
• Show how much time is consumed
• Recommend performance improvements

Features:

• Continuous profiling for Java, Python
• Identifies expensive code paths
• Recommendations for optimization
• Integration with Lambda and EC2
• Flame graphs, heat maps

Use cases:

• Find CPU-intensive code
• Identify memory leaks
• Optimize resource usage
• Reduce latency

Lambda Power Tuning

How to determine minimum memory and compute power:

• Manually run tests, OR

• Use AWS Lambda Power Tuning tool to automate process and graph results

• Open-source tool by Alex Casalboni

• Tests function with different memory settings

• Outputs optimal memory/cost/duration balance

• Step Functions-based workflow

128MB → Duration: 2000ms, Cost: $0.000004
256MB → Duration: 1000ms, Cost: $0.000004
512MB → Duration: 500ms,  Cost: $0.000004  ← optimal
1024MB → Duration: 480ms, Cost: $0.000008

AWS Compute Optimizer

What it does:

• Identify optimal resource configurations
• EC2 instance types
• EBS volume configurations
• Lambda function memory sizes

How it works:

• Analyzes historical utilization metrics
• Machine learning recommendations
• Cost and performance trade-offs

Lambda REPORT Line

REPORT RequestId: xxx
Duration: 250.00 ms
Billed Duration: 300 ms
Memory Size: 256 MB
Max Memory Used: 240 MB    ← Close! Increase memory
Init Duration: 500.00 ms   ← Cold start

X-Ray for Performance

• Trace latency per service
• Identify slow subsegments
• Compare traces over time
• Service Map shows bottlenecks