optimization-benchmark（优化基准）

Benchmark Suite Skill

Overview

This skill provides comprehensive automated performance testing capabilities including benchmark execution, regression detection, performance validation, and quality assessment for ensuring optimal system performance.

When to Use

• Running performance benchmark suites before deployment
• Detecting performance regressions between versions
• Validating SLA compliance through automated testing
• Load, stress, and endurance testing
• CI/CD pipeline performance quality gates
• Comparing performance across configurations

Quick Start

# Run comprehensive benchmark suite
npx claude-flow benchmark-run --suite comprehensive --duration 300

# Execute specific benchmark
npx claude-flow benchmark-run --suite throughput --iterations 10

# Compare with baseline
npx claude-flow benchmark-compare --current <results> --baseline <baseline>

# Quality assessment
npx claude-flow quality-assess --target swarm-performance --criteria throughput,latency

# Performance validation
npx claude-flow validate-performance --results <file> --criteria <file>

Architecture

+-----------------------------------------------------------+
|                   Benchmark Suite                          |
+-----------------------------------------------------------+
|  Benchmark Runner  |  Regression Detector  |  Validator   |
+--------------------+-----------------------+--------------+
         |                     |                    |
         v                     v                    v
+------------------+  +-------------------+  +--------------+
| Benchmark Types  |  | Detection Methods |  | Validation   |
| - Throughput     |  | - Statistical     |  | - SLA        |
| - Latency        |  | - ML-based        |  | - Regression |
| - Scalability    |  | - Threshold       |  | - Scalability|
| - Coordination   |  | - Trend Analysis  |  | - Reliability|
+------------------+  +-------------------+  +--------------+
         |                     |                    |
         v                     v                    v
+-----------------------------------------------------------+
|              Reporter & Comparator                         |
+-----------------------------------------------------------+

Benchmark Types

Standard Benchmarks

Test Campaign Types

1. Load Testing: Gradual ramp-up to sustained load
2. Stress Testing: Find breaking points
3. Volume Testing: Large data set handling
4. Endurance Testing: Long-duration stability
5. Spike Testing: Sudden load changes
6. Configuration Testing: Different settings comparison

Core Capabilities

1. Comprehensive Benchmarking

// Run benchmark suite
const results = await benchmarkSuite.run({
  duration: 300000,      // 5 minutes
  iterations: 10,        // 10 iterations
  warmupTime: 30000,     // 30 seconds warmup
  cooldownTime: 10000,   // 10 seconds cooldown
  parallel: false,       // Sequential execution
  baseline: previousRun  // Compare with baseline
});

// Results include:
// - summary: Overall scores and status
// - detailed: Per-benchmark results
// - baseline_comparison: Delta from baseline
// - recommendations: Optimization suggestions

2. Regression Detection

Multi-algorithm detection:

# Detect performance regressions
npx claude-flow detect-regression --current <results> --historical <data>

# Set up automated regression monitoring
npx claude-flow regression-monitor --enable --sensitivity 0.95

3. Automated Performance Testing

// Execute test campaign
const campaign = await tester.runTestCampaign({
  tests: [
    { type: 'load', config: loadTestConfig },
    { type: 'stress', config: stressTestConfig },
    { type: 'endurance', config: enduranceConfig }
  ],
  constraints: {
    maxDuration: 3600000,  // 1 hour max
    failFast: true         // Stop on first failure
  }
});

4. Performance Validation

Validation framework with multi-criteria assessment:

MCP Integration

// Comprehensive benchmark integration
const benchmarkIntegration = {
  // Execute performance benchmarks
  async runBenchmarks(config = {}) {
    const [benchmark, metrics, trends, cost] = await Promise.all([
      mcp.benchmark_run({ suite: config.suite || 'comprehensive' }),
      mcp.metrics_collect({ components: ['system', 'agents', 'coordination'] }),
      mcp.trend_analysis({ metric: 'performance', period: '24h' }),
      mcp.cost_analysis({ timeframe: '24h' })
    ]);

    return { benchmark, metrics, trends, cost, timestamp: Date.now() };
  },

  // Quality assessment
  async assessQuality(criteria) {
    return await mcp.quality_assess({
      target: 'swarm-performance',
      criteria: criteria || ['throughput', 'latency', 'reliability', 'scalability']
    });
  }
};

Key Metrics

Benchmark Targets

const benchmarkTargets = {
  throughput: {
    requests_per_second: { min: 1000, optimal: 5000 },
    tasks_per_second: { min: 100, optimal: 500 },
    messages_per_second: { min: 10000, optimal: 50000 }
  },
  latency: {
    p50: { max: 100 },   // 100ms
    p90: { max: 200 },   // 200ms
    p95: { max: 500 },   // 500ms
    p99: { max: 1000 },  // 1s
    max: { max: 5000 }   // 5s
  },
  scalability: {
    linear_coefficient: { min: 0.8 },
    efficiency_retention: { min: 0.7 }
  }
};

CI/CD Quality Gates

Load Testing Example

// Load test with gradual ramp-up
const loadTest = {
  type: 'load',
  phases: [
    { phase: 'ramp-up', duration: 60000, startLoad: 10, endLoad: 100 },
    { phase: 'sustained', duration: 300000, load: 100 },
    { phase: 'ramp-down', duration: 30000, startLoad: 100, endLoad: 0 }
  ],
  successCriteria: {
    p99_latency: { max: 1000 },
    error_rate: { max: 0.01 },
    throughput: { min: 80 }  // % of expected
  }
};

Stress Testing Example

// Stress test to find breaking point
const stressTest = {
  type: 'stress',
  startLoad: 100,
  maxLoad: 10000,
  loadIncrement: 100,
  duration: 60000,  // Per load level
  breakingCriteria: {
    error_rate: { max: 0.05 },    // 5% errors
    latency_p99: { max: 5000 },   // 5s latency
    timeout_rate: { max: 0.10 }   // 10% timeouts
  }
};

Integration Points

Best Practices

1. Consistent Environment: Run benchmarks in consistent, isolated environments
2. Warmup Period: Always include warmup to eliminate cold-start effects
3. Multiple Iterations: Run multiple iterations for statistical significance
4. Baseline Maintenance: Keep baseline updated with expected performance
5. Historical Tracking: Store all benchmark results for trend analysis
6. Realistic Workloads: Use production-like workload patterns

Example: CI/CD Integration

#!/bin/bash
# ci-performance-gate.sh

# Run benchmark suite
RESULTS=$(npx claude-flow benchmark-run --suite quick --output json)

# Compare with baseline
COMPARISON=$(npx claude-flow benchmark-compare \
  --current "$RESULTS" \
  --baseline ./baseline.json)

# Check for regressions
if echo "$COMPARISON" | jq -e '.regression_detected == true' > /dev/null; then
  echo "Performance regression detected!"
  echo "$COMPARISON" | jq '.regressions'
  exit 1
fi

echo "Performance validation passed"
exit 0

Related Skills

• optimization-monitor - Real-time performance monitoring
• optimization-analyzer - Bottleneck analysis and reporting
• optimization-load-balancer - Load distribution optimization
• optimization-topology - Topology performance testing

Version History

• 1.0.0 (2026-01-02): Initial release - converted from benchmark-suite agent with comprehensive benchmarking, regression detection, automated testing, and performance validation