# curl timing breakdown
cat curl-format.txt
time_namelookup:    %{time_namelookup}s\n
time_connect:       %{time_connect}s\n
time_appconnect:    %{time_appconnect}s\n   # TLS done
time_pretransfer:   %{time_pretransfer}s\n
time_starttransfer: %{time_starttransfer}s\n # TTFB
time_total:         %{time_total}s\n

# Response có các header này → CDN từ chối cache
Cache-Control: no-cache, no-store
Set-Cookie: session=abc123   ← CloudFront mặc định không cache response có Set-Cookie
Vary: Cookie, Authorization  ← quá nhiều variation → cache không hiệu quả

# Static assets (JS, CSS, images)
Cache-Control: public, max-age=31536000, immutable
 
# API response có thể cache (không personalised)
Cache-Control: public, s-maxage=60, stale-while-revalidate=30
 
# Dynamic nhưng authenticated → không cache trên CDN
Cache-Control: private, no-store

# CloudFront cache policy — tách cookie ra khỏi cache key
CachePolicy:
  Name: "api-cache-policy"
  DefaultTTL: 60
  CookiesConfig:
    CookieBehavior: none   # không đưa cookie vào cache key
  HeadersConfig:
    HeaderBehavior: whitelist
    Headers:
      - Accept-Language    # chỉ vary theo những header cần thiết

User (US) → Server (Singapore) → 3rd API call 1 → 3rd API call 2 → 3rd API call 3
           ←──────────────────── 150ms ──────────── 150ms ─────── 150ms ──────────
Total: 150ms RTT × 3 hops = 450ms chỉ do mạng, chưa tính processing

// BAD: serial calls — mỗi call chờ call trước
@GetMapping("/dashboard")
public DashboardResponse getDashboard(@RequestHeader("userId") Long userId) {
    UserProfile profile = userService.getProfile(userId);         // 150ms
    List<Order> orders = orderService.getRecentOrders(userId);    // 150ms
    AccountBalance balance = paymentService.getBalance(userId);   // 150ms
    List<Notification> notifs = notifService.getUnread(userId);   // 150ms
    // Total: 600ms chỉ từ serial calls
    return DashboardResponse.of(profile, orders, balance, notifs);
}
 
// GOOD: parallel với CompletableFuture
@GetMapping("/dashboard")
public DashboardResponse getDashboard(@RequestHeader("userId") Long userId) {
    CompletableFuture<UserProfile> profileF =
        CompletableFuture.supplyAsync(() -> userService.getProfile(userId));
    CompletableFuture<List<Order>> ordersF =
        CompletableFuture.supplyAsync(() -> orderService.getRecentOrders(userId));
    CompletableFuture<AccountBalance> balanceF =
        CompletableFuture.supplyAsync(() -> paymentService.getBalance(userId));
    CompletableFuture<List<Notification>> notifsF =
        CompletableFuture.supplyAsync(() -> notifService.getUnread(userId));
 
    CompletableFuture.allOf(profileF, ordersF, balanceF, notifsF).join();
    // Total: ~150ms (chạy song song, bottleneck là call chậm nhất)
    return DashboardResponse.of(
        profileF.join(), ordersF.join(), balanceF.join(), notifsF.join());
}

Resource 1: TCP(150ms) + TLS(300ms) + Transfer = 470ms
Resource 2: TCP(150ms) + TLS(300ms) + Transfer = 470ms
Resource 3: TCP(150ms) + TLS(300ms) + Transfer = 470ms
→ 3 resources song song: ~470ms; nối tiếp: ~1.4s

# Nginx — bật HTTP/2
server {
    listen 443 ssl http2;
    ssl_protocols TLSv1.3;          # 1-RTT handshake
    ssl_session_cache shared:SSL:10m;  # session reuse → 0-RTT resumption
    ssl_session_timeout 1d;
}

                    ┌─────────────────────────────────────┐
                    │           Route 53                  │
                    │    Latency-Based Routing            │
                    └──────────┬──────────────────────────┘
                               │
              ┌────────────────┴───────────────┐
              │                                │
   ┌──────────▼──────────┐          ┌──────────▼──────────┐
   │   EKS ap-southeast-1│          │   EKS us-east-1     │
   │   (Singapore)       │          │   (Virginia)        │
   │   60ms cho SG users │          │   60ms cho US users │
   └──────────┬──────────┘          └──────────┬──────────┘
              │                                │
   ┌──────────▼──────────┐          ┌──────────▼──────────┐
   │  RDS ap-southeast-1 │◄─────────│  RDS us-east-1      │
   │  (Primary)          │ Replication  (Read Replica)    │
   └─────────────────────┘          └─────────────────────┘

# Terraform — Route 53 Latency Routing
resource "aws_route53_record" "api_sg" {
  zone_id        = var.zone_id
  name           = "api.example.com"
  type           = "A"
  set_identifier = "singapore"
 
  latency_routing_policy {
    region = "ap-southeast-1"
  }
 
  alias {
    name    = aws_lb.alb_sg.dns_name
    zone_id = aws_lb.alb_sg.zone_id
    evaluate_target_health = true
  }
}
 
resource "aws_route53_record" "api_us" {
  zone_id        = var.zone_id
  name           = "api.example.com"
  type           = "A"
  set_identifier = "us-east"
 
  latency_routing_policy {
    region = "us-east-1"
  }
 
  alias {
    name    = aws_lb.alb_us.dns_name
    zone_id = aws_lb.alb_us.zone_id
    evaluate_target_health = true
  }
}

□ 1. Đo waterfall từ US location (WebPageTest / curl timing)
□ 2. Check TTFB: > 500ms → vấn đề backend/network, không phải frontend
□ 3. Check X-Cache header: "Miss" → fix CDN cache
□ 4. Check Cache-Control response header: no-cache? Set-Cookie?
□ 5. Đếm số serial network calls trong 1 request: N calls × 150ms?
□ 6. Check HTTP version: HTTP/1.1 với nhiều resources → upgrade HTTP/2
□ 7. Check TLS version: TLS 1.2 → 2 RTTs; upgrade TLS 1.3 → 1 RTT
□ 8. Check DNS TTL: thấp quá → nhiều DNS lookup tốn thêm latency
□ 9. Check origin location: server chỉ ở Singapore? → cần multi-region
□ 10. Check static asset CDN: JS/CSS/images đi qua CloudFront chưa?

# CloudFront cache hit ratio theo region
aws cloudwatch get-metric-statistics \
  --namespace AWS/CloudFront \
  --metric-name CacheHitRate \
  --dimensions Name=DistributionId,Value=EXDISTID \
  --start-time 2026-04-08T00:00:00Z \
  --end-time 2026-04-08T23:59:59Z \
  --period 3600 \
  --statistics Average
 
# ALB target response time per AZ (proxy cho per-region)
aws cloudwatch get-metric-statistics \
  --namespace AWS/ApplicationELB \
  --metric-name TargetResponseTime \
  --dimensions Name=LoadBalancer,Value=... \
  --statistics p95 p99

# Prometheus alert — latency cao theo region
- alert: HighLatencyUSRegion
  expr: |
    histogram_quantile(0.95,
      rate(http_request_duration_seconds_bucket{region="us-east-1"}[5m])
    ) > 2
  for: 5m
  annotations:
    summary: "p95 latency US-East = {{ $value }}s — kiểm tra CDN cache hit ratio"
    runbook: "https://wiki/runbooks/geo-latency-debug"