Skip to content

Instantly share code, notes, and snippets.

@akimabs

akimabs/bench.md

Last active November 2, 2025 17:09
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save akimabs/9e4f3f0177cbd66c85a1986bde5b5fcf to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save akimabs/9e4f3f0177cbd66c85a1986bde5b5fcf to your computer and use it in GitHub Desktop.
Download ZIP
benchmark
Raw
bench.md

Hasil Benchmark: WebFlux vs Spring MVC + Virtual Thread

Tanggal: 2 November 2025 Durasi Test: ~3 menit 27 detik Tool: k6 (Concurrent Load Test)

Ringkasan Eksekutif

WebFlux secara signifikan lebih unggul dibanding Spring MVC + Virtual Thread dalam concurrent load test ini:

  • 53% throughput lebih tinggi (9,777 req/s vs 6,393 req/s)
  • 35% lebih cepat (7.18ms vs 11.01ms rata-rata)
  • 40% lebih rendah P95 latency (14.83ms vs 24.69ms)
  • 53% lebih banyak request (880,614 vs 575,379)

Pemenang: WebFlux (Reactive) 🏆

Konfigurasi Test

Pola Load (Per Service)

Ramp up:    15 detik → 50 concurrent users
Hold:       30 detik → 100 concurrent users
Peak:       15 detik → 150 concurrent users
Ramp down:  30 detik → 0 users
Total:      ~90 detik per service

Strategi Test

  • Sequential Isolated Testing - Setiap service ditest terpisah untuk menghindari contention database
  • Cooldown 10 detik antar test agar database settling
  • Fair comparison - Kedua service dapat full database resources

Environment

  • Java Version: 21
  • Spring Boot: 3.5.7
  • Database: Shared (tested sequentially)
  • Max Concurrent Users: 150 VUs

Perbandingan Performance Metrics

1. Response Time (Semakin Rendah Semakin Bagus)

Metric WebFlux Spring MVC Pemenang Peningkatan
Average 7.18ms 11.01ms WebFlux 35% lebih cepat
Median (P50) 6.51ms 9.70ms WebFlux 33% lebih cepat
P90 12.79ms 21.11ms WebFlux 39% lebih cepat
P95 14.83ms 24.69ms WebFlux 40% lebih cepat
Min 0.44ms 0.48ms WebFlux 8% lebih cepat
Max 130.19ms 140.96ms WebFlux 8% lebih cepat

2. Throughput (Semakin Tinggi Semakin Bagus)

Metric WebFlux Spring MVC Pemenang Peningkatan
Requests/detik 9,776.97 6,392.97 WebFlux +53% lebih tinggi
Total Requests 880,614 575,379 WebFlux +53% lebih banyak
Iterations/detik 9,776.97 6,392.97 WebFlux +53% lebih tinggi

3. Reliability

Metric WebFlux Spring MVC Status
Success Rate 100% 100% Keduanya sempurna ✓
Failed Requests 0 0 Keduanya sempurna ✓
Error Rate 0.00% 0.00% Keduanya sempurna ✓

4. Network Performance

Metric WebFlux Spring MVC Pemenang
Data Diterima 130 MB (1.4 MB/s) 164 MB (1.8 MB/s) MVC (+26%)
Data Dikirim 80 MB (890 KB/s) 52 MB (582 KB/s) WebFlux (+54%)
Req Blocked (avg) 0.90µs 2.67µs WebFlux (-66%)
Req Receiving (avg) 12.07µs 45.63µs WebFlux (-74%)
Req Sending (avg) 4.12µs 4.91µs WebFlux (-16%)

Analisis Penggunaan Resource

Profil Resource WebFlux

File: app_resource_20251102_235058.log

Fase CPU % Memory (MB) Threads Open Files
Idle 0.0% 69-74 44 150
Startup 47.7-133% 75-96 15 18-41
Ramp Up 108-145% 96-138 15-16 41-118
Steady State 110-139% 138-194 16 118-167
Peak Load 109-145% 194-280 16 167
Cooldown 56.3% → 0% 280 → 231 16 → 51 169 → 160

Observasi Kunci:

  • Thread count rendah: 15-16 threads (reactive, non-blocking)
  • CPU optimal: 108-145% saat load (1-1.5 cores)
  • Memory growth: 69 MB → 280 MB saat load (lebih tinggi dari sebelumnya)
  • File descriptors: 18 → 167 (stabil di peak)

Profil Resource Spring MVC + Virtual Thread

File: app_resource_20251102_235242.log

Fase CPU % Memory (MB) Threads Open Files
Idle 0.0% 35 40 101
Startup 52-135% 78-105 15 20-62
Ramp Up 103-136% 105-152 15 62-138
Steady State 112-135% 152-191 15 138-167
Peak Load 91-132% 191-207 15 166-168
Cooldown 75.5% → 0% 207 → 195 15 → 49 168 → 101

Observasi Kunci:

  • Thread count rendah: 15 threads (virtual threads enabled!)
  • CPU moderate: 91-136% saat load (1-1.4 cores)
  • Memory stabil: 35 MB → 207 MB saat load
  • File descriptors: 20 → 168 (similar pattern dengan WebFlux)

Perbandingan Resource

Penggunaan CPU

Metric WebFlux Spring MVC Analisis
Peak CPU 145.8% 135.7% WebFlux +7% lebih tinggi tapi menghasilkan +53% throughput
Avg CPU (load) ~120% ~120% Keduanya similar CPU usage (~1 core)
CPU Efficiency Sangat efisien Efisien WebFlux +53% lebih banyak request per CPU cycle

Pemenang: WebFlux - Efisiensi CPU lebih baik = throughput +53% lebih tinggi dengan CPU similar

Penggunaan Memory

Metric WebFlux Spring MVC Pemenang
Idle Memory 69 MB 35 MB MVC (baseline -49%)
Peak Memory 280 MB 207 MB MVC (-26% lebih sedikit)
Memory Growth 69 → 280 MB (+306%) 35 → 207 MB (+491%) WebFlux (growth lebih stabil)

Pemenang: Spring MVC - Hemat 26% memory di peak load

Jumlah Thread

Metric WebFlux Spring MVC Analisis
Idle Threads 44 40 Baseline similar
Active Threads 15-16 15 SAMA! Keduanya pakai 15 threads
Thread Model Event loop (reactive) Virtual threads (blocking) Paradigma berbeda

Pemenang: TIE - Thread count identik (virtual threads tidak terlihat di metrics)

Open File Descriptors (Connections)

Metric WebFlux Spring MVC Pemenang
Idle 150 101 MVC (lebih rendah)
Peak 167 168 TIE (hampir identik)
Growth 18 → 167 (9x) 20 → 168 (8.4x) Similar growth pattern

Pemenang: TIE - Connection handling serupa

Analisis Timeline Detail

Timeline WebFlux (0:00 - 1:30)

00:00 - 00:03   Idle          CPU: 0%        Memory: 69-74 MB Threads: 44
00:03 - 00:13   App Start     CPU: 47-133%   Memory: 75-96    Threads: 15
00:13 - 00:43   Ramp Up       CPU: 108-145%  Memory: 96-138   Threads: 15-16
00:43 - 01:13   Steady 100VU  CPU: 110-139%  Memory: 138-194  Threads: 16
01:13 - 01:28   Peak 150VU    CPU: 109-145%  Memory: 194-280  Threads: 16
01:28 - 01:35   Shutdown      CPU: 56→0%     Memory: 280→231  Threads: 16→51

Observasi: CPU optimal di 108-145% (1-1.5 cores) - sangat efisien!

Timeline Spring MVC (1:40 - 3:05)

01:40 - 01:44   Idle          CPU: 0%        Memory: 35 MB    Threads: 40
01:44 - 01:54   App Start     CPU: 52-135%   Memory: 78-105   Threads: 15
01:54 - 02:24   Ramp Up       CPU: 103-136%  Memory: 105-152  Threads: 15
02:24 - 02:54   Steady 100VU  CPU: 112-135%  Memory: 152-191  Threads: 15
02:54 - 03:09   Peak 150VU    CPU: 91-132%   Memory: 191-207  Threads: 15
03:09 - 03:16   Shutdown      CPU: 75.5→0%   Memory: 207→195  Threads: 15→49

Observasi: CPU moderate di 91-136% (1-1.4 cores) - tidak fully utilize multi-core

Visualisasi Performance Summary

Distribusi Response Time

WebFlux:
Min ━━━━━━━━━━━ Max
0.44ms  6.51ms (P50)  14.83ms (P95)                                  130ms

MVC:
Min ━━━━━━━━━━━━━━━━━━━━ Max
0.48ms     9.70ms (P50)     24.69ms (P95)                            141ms

Visualisasi: WebFlux 33-40% lebih cepat di semua percentile

Throughput Over Time

WebFlux:     ████████████████████████████████████████████████████████████████ 9,777 req/s
Spring MVC:  ████████████████████████████████████████████ 6,393 req/s

Visualisasi: WebFlux handle +53% lebih banyak request

Rangkuman Raw Data

k6 Output - WebFlux

checks.........................: 100.00% ✓ 2641842     ✗ 0
http_req_duration..............: avg=7.18ms  min=438µs   med=6.51ms max=130.19ms p(90)=12.79ms p(95)=14.83ms
http_req_failed................: 0.00%   ✓ 0           ✗ 880614
http_reqs......................: 880614  9776.966036/s
iterations.....................: 880614  9776.966036/s
vus............................: 1       min=1         max=149
vus_max........................: 150     min=150       max=150

k6 Output - Spring MVC + Virtual Thread

checks.........................: 100.00% ✓ 1726137     ✗ 0
http_req_duration..............: avg=11.01ms min=484µs    med=9.7ms  max=140.96ms p(90)=21.11ms p(95)=24.69ms
http_req_failed................: 0.00%   ✓ 0           ✗ 575379
http_reqs......................: 575379  6392.972496/s
iterations.....................: 575379  6392.972496/s
vus............................: 1       min=1         max=149
vus_max........................: 150     min=150       max=150

Rekomendasi Final

TETAP di WebFlux dengan optimizations:

  • ✅ Performance sudah optimal (9,777 req/s)
  • ✅ Latency rendah (7.18ms average, 14.83ms P95)
  • ✅ CPU efficiency tinggi (+53% vs MVC)
  • ✅ Codebase sudah mature
  • ✅ Tidak perlu effort migrasi (save 18-25 jam development)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment