返回 Skill 列表
extension
分类: 开发与工程无需 API Key

performance-benchmark-suite

SDK性能基准测试和回归检测

person作者: jakexiaohubgithub

Performance Benchmark Suite Skill

Overview

This skill implements comprehensive SDK performance benchmarking, tracking latency, throughput, memory usage, and detecting performance regressions across versions.

Capabilities

  • Measure latency percentiles (p50, p95, p99)
  • Track memory usage and allocation patterns
  • Detect performance regressions automatically
  • Generate visual benchmark reports
  • Compare performance across SDK versions
  • Implement microbenchmarks for critical paths
  • Configure continuous benchmarking in CI
  • Support load testing scenarios

Target Processes

  • Performance Benchmarking
  • SDK Testing Strategy
  • SDK Versioning and Release Management

Integration Points

  • k6 for load testing
  • Artillery for HTTP benchmarking
  • hyperfine for CLI benchmarking
  • Benchmark.js for JavaScript
  • pytest-benchmark for Python
  • Continuous benchmark systems (Bencher)

Input Requirements

  • Performance requirements (SLOs)
  • Benchmark scenarios
  • Baseline versions for comparison
  • Environment specifications
  • Reporting requirements

Output Artifacts

  • Benchmark test suite
  • Performance baseline data
  • Regression detection rules
  • Visual benchmark reports
  • CI benchmark configuration
  • Historical trend analysis

Usage Example

skill:
  name: performance-benchmark-suite
  context:
    tool: k6
    scenarios:
      - name: basic-crud
        operations: ["create", "read", "update", "delete"]
        vus: 10
        duration: "30s"
      - name: high-load
        vus: 100
        duration: "5m"
    slos:
      p95_latency: "100ms"
      p99_latency: "500ms"
      error_rate: "0.1%"
    compareWith: "v1.0.0"
    regressionThreshold: "10%"

Best Practices

  1. Establish baselines before optimization
  2. Track percentiles, not just averages
  3. Run benchmarks in consistent environments
  4. Automate regression detection in CI
  5. Monitor memory alongside latency
  6. Document benchmark methodology