hotime/example/benchmark/extreme_test.go

package main

import (
	"encoding/json"
	"fmt"
	"io"
	"net/http"
	"runtime"
	"sync"
	"sync/atomic"
	"time"
)

// 压测配置
type BenchConfig struct {
	URL         string
	Concurrency int
	Duration    time.Duration
	Timeout     time.Duration
}

// 压测结果
type BenchResult struct {
	TotalRequests   int64
	SuccessRequests int64
	FailedRequests  int64
	TotalDuration   time.Duration
	MinLatency      int64 // 纳秒
	MaxLatency      int64
	AvgLatency      int64
	P50Latency      int64 // 50分位
	P90Latency      int64 // 90分位
	P99Latency      int64 // 99分位
	QPS             float64
}

// 延迟收集器
type LatencyCollector struct {
	latencies []int64
	mu        sync.Mutex
}

func (lc *LatencyCollector) Add(latency int64) {
	lc.mu.Lock()
	lc.latencies = append(lc.latencies, latency)
	lc.mu.Unlock()
}

func (lc *LatencyCollector) GetPercentile(p float64) int64 {
	lc.mu.Lock()
	defer lc.mu.Unlock()

	if len(lc.latencies) == 0 {
		return 0
	}

	// 简单排序取百分位
	n := len(lc.latencies)
	idx := int(float64(n) * p)
	if idx >= n {
		idx = n - 1
	}

	// 部分排序找第idx个元素
	return quickSelect(lc.latencies, idx)
}

func quickSelect(arr []int64, k int) int64 {
	if len(arr) == 1 {
		return arr[0]
	}

	pivot := arr[len(arr)/2]
	var left, right, equal []int64

	for _, v := range arr {
		if v < pivot {
			left = append(left, v)
		} else if v > pivot {
			right = append(right, v)
		} else {
			equal = append(equal, v)
		}
	}

	if k < len(left) {
		return quickSelect(left, k)
	} else if k < len(left)+len(equal) {
		return pivot
	}
	return quickSelect(right, k-len(left)-len(equal))
}

func main() {
	// 最大化利用CPU
	runtime.GOMAXPROCS(runtime.NumCPU())

	fmt.Println("==========================================")
	fmt.Println("  🔥 HoTime 极限压力测试 🔥")
	fmt.Println("==========================================")
	fmt.Printf("CPU 核心数: %d\n", runtime.NumCPU())
	fmt.Println()

	// 极限测试配置
	configs := []BenchConfig{
		{URL: "http://127.0.0.1:8081/app/test/hello", Concurrency: 500, Duration: 15 * time.Second, Timeout: 10 * time.Second},
		{URL: "http://127.0.0.1:8081/app/test/hello", Concurrency: 1000, Duration: 15 * time.Second, Timeout: 10 * time.Second},
		{URL: "http://127.0.0.1:8081/app/test/hello", Concurrency: 2000, Duration: 15 * time.Second, Timeout: 10 * time.Second},
		{URL: "http://127.0.0.1:8081/app/test/hello", Concurrency: 5000, Duration: 15 * time.Second, Timeout: 10 * time.Second},
		{URL: "http://127.0.0.1:8081/app/test/hello", Concurrency: 10000, Duration: 15 * time.Second, Timeout: 10 * time.Second},
	}

	// 检查服务
	fmt.Println("正在检查服务是否可用...")
	if !checkService(configs[0].URL) {
		fmt.Println("❌ 服务不可用，请先启动示例应用")
		return
	}
	fmt.Println("✅ 服务已就绪")
	fmt.Println()

	// 预热
	fmt.Println("🔄 预热中 (5秒)...")
	warmup(configs[0].URL, 100, 5*time.Second)
	fmt.Println("✅ 预热完成")
	fmt.Println()

	var maxQPS float64
	var maxConcurrency int

	// 执行极限测试
	for i, config := range configs {
		fmt.Printf("══════════════════════════════════════════\n")
		fmt.Printf("【极限测试 %d】并发数: %d, 持续时间: %v\n", i+1, config.Concurrency, config.Duration)
		fmt.Printf("══════════════════════════════════════════\n")

		result := runBenchmark(config)
		printResult(result)

		if result.QPS > maxQPS {
			maxQPS = result.QPS
			maxConcurrency = config.Concurrency
		}

		// 检查是否达到瓶颈
		successRate := float64(result.SuccessRequests) / float64(result.TotalRequests) * 100
		if successRate < 95 {
			fmt.Println("\n⚠️ 成功率低于95%，已达到服务极限！")
			break
		}

		if result.AvgLatency > int64(100*time.Millisecond) {
			fmt.Println("\n⚠️ 平均延迟超过100ms，已达到服务极限！")
			break
		}

		fmt.Println()

		// 测试间隔
		if i < len(configs)-1 {
			fmt.Println("冷却 5 秒...")
			time.Sleep(5 * time.Second)
			fmt.Println()
		}
	}

	// 最终报告
	fmt.Println()
	fmt.Println("══════════════════════════════════════════")
	fmt.Println("  📊 极限测试总结")
	fmt.Println("══════════════════════════════════════════")
	fmt.Printf("最高 QPS:      %.2f 请求/秒\n", maxQPS)
	fmt.Printf("最佳并发数:    %d\n", maxConcurrency)
	fmt.Println()

	// 并发用户估算
	estimateUsers(maxQPS)
}

func checkService(url string) bool {
	client := &http.Client{Timeout: 3 * time.Second}
	resp, err := client.Get(url)
	if err != nil {
		return false
	}
	defer resp.Body.Close()
	return resp.StatusCode == 200
}

func warmup(url string, concurrency int, duration time.Duration) {
	client := &http.Client{
		Timeout: 5 * time.Second,
		Transport: &http.Transport{
			MaxIdleConns:        concurrency * 2,
			MaxIdleConnsPerHost: concurrency * 2,
		},
	}

	done := make(chan struct{})
	var wg sync.WaitGroup

	for i := 0; i < concurrency; i++ {
		wg.Add(1)
		go func() {
			defer wg.Done()
			for {
				select {
				case <-done:
					return
				default:
					resp, err := client.Get(url)
					if err == nil {
						io.Copy(io.Discard, resp.Body)
						resp.Body.Close()
					}
				}
			}
		}()
	}

	time.Sleep(duration)
	close(done)
	wg.Wait()
}

func runBenchmark(config BenchConfig) BenchResult {
	var (
		totalRequests   int64
		successRequests int64
		failedRequests  int64
		totalLatency    int64
		minLatency      int64 = int64(time.Hour)
		maxLatency      int64
		mu              sync.Mutex
	)

	collector := &LatencyCollector{
		latencies: make([]int64, 0, 100000),
	}

	// 高性能HTTP客户端
	client := &http.Client{
		Timeout: config.Timeout,
		Transport: &http.Transport{
			MaxIdleConns:        config.Concurrency * 2,
			MaxIdleConnsPerHost: config.Concurrency * 2,
			MaxConnsPerHost:     config.Concurrency * 2,
			IdleConnTimeout:     90 * time.Second,
			DisableKeepAlives:   false,
			DisableCompression:  true,
		},
	}

	done := make(chan struct{})
	var wg sync.WaitGroup

	startTime := time.Now()

	// 启动并发
	for i := 0; i < config.Concurrency; i++ {
		wg.Add(1)
		go func() {
			defer wg.Done()

			for {
				select {
				case <-done:
					return
				default:
					reqStart := time.Now()
					success := makeRequest(client, config.URL)
					latency := time.Since(reqStart).Nanoseconds()

					atomic.AddInt64(&totalRequests, 1)
					atomic.AddInt64(&totalLatency, latency)

					if success {
						atomic.AddInt64(&successRequests, 1)
					} else {
						atomic.AddInt64(&failedRequests, 1)
					}

					// 采样收集延迟（每100个请求采样1个，减少内存开销）
					if atomic.LoadInt64(&totalRequests)%100 == 0 {
						collector.Add(latency)
					}

					mu.Lock()
					if latency < minLatency {
						minLatency = latency
					}
					if latency > maxLatency {
						maxLatency = latency
					}
					mu.Unlock()
				}
			}
		}()
	}

	time.Sleep(config.Duration)
	close(done)
	wg.Wait()

	totalDuration := time.Since(startTime)

	result := BenchResult{
		TotalRequests:   totalRequests,
		SuccessRequests: successRequests,
		FailedRequests:  failedRequests,
		TotalDuration:   totalDuration,
		MinLatency:      minLatency,
		MaxLatency:      maxLatency,
		P50Latency:      collector.GetPercentile(0.50),
		P90Latency:      collector.GetPercentile(0.90),
		P99Latency:      collector.GetPercentile(0.99),
	}

	if totalRequests > 0 {
		result.AvgLatency = totalLatency / totalRequests
		result.QPS = float64(totalRequests) / totalDuration.Seconds()
	}

	return result
}

func makeRequest(client *http.Client, url string) bool {
	resp, err := client.Get(url)
	if err != nil {
		return false
	}
	defer resp.Body.Close()

	body, err := io.ReadAll(resp.Body)
	if err != nil {
		return false
	}

	if resp.StatusCode != 200 {
		return false
	}

	var result map[string]interface{}
	if err := json.Unmarshal(body, &result); err != nil {
		return false
	}

	if status, ok := result["status"].(float64); !ok || status != 0 {
		return false
	}

	return true
}

func printResult(result BenchResult) {
	successRate := float64(result.SuccessRequests) / float64(result.TotalRequests) * 100

	fmt.Printf("总请求数:     %d\n", result.TotalRequests)
	fmt.Printf("成功请求:     %d\n", result.SuccessRequests)
	fmt.Printf("失败请求:     %d\n", result.FailedRequests)
	fmt.Printf("成功率:       %.2f%%\n", successRate)
	fmt.Printf("总耗时:       %v\n", result.TotalDuration.Round(time.Millisecond))
	fmt.Printf("QPS:          %.2f 请求/秒\n", result.QPS)
	fmt.Println("------------------------------------------")
	fmt.Printf("最小延迟:     %v\n", time.Duration(result.MinLatency).Round(time.Microsecond))
	fmt.Printf("平均延迟:     %v\n", time.Duration(result.AvgLatency).Round(time.Microsecond))
	fmt.Printf("P50延迟:      %v\n", time.Duration(result.P50Latency).Round(time.Microsecond))
	fmt.Printf("P90延迟:      %v\n", time.Duration(result.P90Latency).Round(time.Microsecond))
	fmt.Printf("P99延迟:      %v\n", time.Duration(result.P99Latency).Round(time.Microsecond))
	fmt.Printf("最大延迟:     %v\n", time.Duration(result.MaxLatency).Round(time.Microsecond))

	// 性能评级
	fmt.Print("\n性能评级:     ")
	switch {
	case result.QPS >= 200000:
		fmt.Println("🏆 卓越 (QPS >= 200K)")
	case result.QPS >= 100000:
		fmt.Println("🚀 优秀 (QPS >= 100K)")
	case result.QPS >= 50000:
		fmt.Println("⭐ 良好 (QPS >= 50K)")
	case result.QPS >= 20000:
		fmt.Println("👍 中上 (QPS >= 20K)")
	case result.QPS >= 10000:
		fmt.Println("📊 中等 (QPS >= 10K)")
	default:
		fmt.Println("⚠️ 一般 (QPS < 10K)")
	}
}

func estimateUsers(maxQPS float64) {
	fmt.Println("📈 并发用户数估算（基于不同使用场景）:")
	fmt.Println("------------------------------------------")

	scenarios := []struct {
		name            string
		requestInterval float64 // 用户平均请求间隔（秒）
	}{
		{"高频交互（每秒1次请求）", 1},
		{"活跃用户（每5秒1次请求）", 5},
		{"普通浏览（每10秒1次请求）", 10},
		{"低频访问（每30秒1次请求）", 30},
		{"偶尔访问（每60秒1次请求）", 60},
	}

	for _, s := range scenarios {
		users := maxQPS * s.requestInterval
		fmt.Printf("%-30s ~%d 用户\n", s.name, int(users))
	}

	fmt.Println()
	fmt.Println("💡 实际生产环境建议保留 30-50% 性能余量")
	fmt.Printf("   安全并发用户数: %d - %d (普通浏览场景)\n",
		int(maxQPS*10*0.5), int(maxQPS*10*0.7))
}