Skip to content

Performance Benchmark

Jump to bottom
adriancs edited this page Jun 22, 2026 · 6 revisions

The following stress test was run on the following computer specs:

The source code of running this test is available at: HearthPortableWebServer.StressTest

Target web application in the test: Un-initialized Hearth CMS (Pageless ASP.NET Web Forms CMS)

  • OS: Windows 11 Pro x64 25H2 (OS build: 26200.8655)
  • CPU: Intel Core i7-4770S @ 3.10GHz (Max Turbo Boost: 3.90GHz)
  • Cores: 4 Physical, 8 Logical Processors
  • Cache: L1: 256 KB | L2: 1.0 MB | L3: 8.0 MB
  • RAM: 24GB DDR3
  • Storage: Samsung SSD 870 EVO 500GB

Hearth Portable ASP.NET Web Server - Stress Test Result:

============================================================================================
Portable ASP.NET Web Server - Stress Test Report
============================================================================================
Generated  : 2026-06-21 20:59:24
Target URL : http://localhost:8080/
Duration   : 10 s measured per level (warmup 3 s)
Machine    : ZEN, 8 logical CPUs, x86 test client
--------------------------------------------------------------------------------------------

Per-level results (latency in milliseconds):

Conc   Requests     Req/sec       Req/min       OK   Fail      Avg     p50     p90     p95     p99      Max
----- ---------- ----------- ------------- -------- ------ -------- ------- ------- ------- ------- --------
   1     33,458     3,345.8       200,746   33,458      0     0.30     0.3     0.3     0.3     0.4      3.9
   2     60,884     6,088.2       365,293   60,884      0     0.33     0.3     0.4     0.4     0.7      5.3
   4     98,468     9,846.5       590,791   98,468      0     0.41     0.4     0.4     0.5     1.0     11.1
   8    134,609    13,460.4       807,624  134,609      0     0.59     0.5     0.7     0.9     1.5     83.1
  16    144,943    14,493.5       869,608  144,943      0     1.10     1.0     1.5     1.9     2.6     11.3
  32    147,021    14,700.0       882,001  147,021      0     2.18     2.0     3.0     3.5     4.5     17.9
  64    138,550    13,849.2       830,951  138,550      0     4.62     4.2     8.3     9.7    13.5     88.3
 128    149,914    14,979.9       898,796  149,914      0     8.53     8.3    13.6    15.0    20.6     58.1
 256    127,077    12,695.9       761,756  127,077      0    20.14    18.6    33.1    38.8    48.5     75.2

--------------------------------------------------------------------------------------------
PEAK THROUGHPUT
 Concurrency level : 128
 Requests / second : 14,979.9
 Requests / minute : 898,796
 Avg latency       : 8.53 ms
 p99 latency       : 20.60 ms
--------------------------------------------------------------------------------------------
Status code distribution (all levels):
 200                      : 1,034,924
--------------------------------------------------------------------------------------------
Total requests sent : 1,034,924
Total failures      : 0
Total data received : 3,532.4 MB
============================================================================================

Windows 11 IIS

Windows 11 built-in IIS server, same application, and run the stress test, and here is the result (single worker process):

============================================================================================
Portable ASP.NET Web Server - Stress Test Report
============================================================================================
Generated  : 2026-06-21 21:07:48
Target URL : http://localhost:8006/
Duration   : 10 s measured per level (warmup 3 s)
Machine    : ZEN, 8 logical CPUs, x86 test client
--------------------------------------------------------------------------------------------

Per-level results (latency in milliseconds):

Conc   Requests     Req/sec       Req/min       OK   Fail      Avg     p50     p90     p95     p99      Max
----- ---------- ----------- ------------- -------- ------ -------- ------- ------- ------- ------- --------
   1     28,251     2,825.0       169,500   28,251      0     0.35     0.3     0.4     0.4     0.5      1.6
   2     46,986     4,698.4       281,907   46,986      0     0.42     0.4     0.5     0.5     0.8      3.4
   4     77,771     7,777.1       466,624   77,771      0     0.51     0.5     0.6     0.6     1.2     11.3
   8    107,843    10,782.7       646,962  107,843      0     0.74     0.7     0.9     1.1     1.9     25.4
  16    123,504    12,348.9       740,935  123,504      0     1.29     1.2     1.7     2.1     2.8     15.6
  32    124,587    12,456.1       747,366  124,587      0     2.57     2.4     3.4     3.9     4.9     17.8
  64    126,916    12,683.4       761,003  126,916      0     5.04     4.9     7.1     8.0     9.2     19.2
 128    128,959    12,881.1       772,866  128,959      0     9.93    10.0    14.6    15.7    17.8     71.0
 256    122,203    12,208.0       732,480  122,203      0    20.95    20.6    28.5    31.3    36.6     58.7

--------------------------------------------------------------------------------------------
PEAK THROUGHPUT
 Concurrency level : 128
 Requests / second : 12,881.1
 Requests / minute : 772,866
 Avg latency       : 9.93 ms
 p99 latency       : 17.83 ms
--------------------------------------------------------------------------------------------
Status code distribution (all levels):
 200                      : 887,020
--------------------------------------------------------------------------------------------
Total requests sent : 887,020
Total failures      : 0
Total data received : 3,027.6 MB
============================================================================================

Head-to-head (req/sec)

Concurrency Portable server IIS (1 worker) Portable advantage
1 3,346 2,825 +18%
2 6,088 4,698 +30%
4 9,847 7,777 +27%
8 13,460 10,783 +25%
16 14,494 12,349 +17%
32 14,700 12,456 +18%
64 13,849 12,683 +9%
128 14,980 12,881 +16%
256 12,696 12,208 +4%

Analysis

  • Peak Performance: The portable server maxed out at ~14,980 req/s vs IIS at ~12,881 req/s—clocking in about 16% faster overall, and leading at every single concurrency level.
  • Latency: The portable server maintained lower latency across the board (e.g., at Concurrency 8: 0.59 ms vs 0.74 ms).
  • Stability: Both servers successfully processed 100% of requests with zero failures.

Why is the portable server faster here?

This does not automatically mean the Hearth portable server is inherently "better" than IIS. Rather, it indicates the portable server executes fewer instructions per request. In a CPU-bound loopback test, this minimalism translates directly into higher throughput.

  1. Shorter Pipeline: Both servers share the same kernel driver (HTTP.sys), which explains why their base numbers are in the same ballpark. However, above the kernel layer, IIS executes a heavy native module pipeline (Authentication, Request Filtering, Static/Dynamic Content, Compression, Health Monitoring, etc.). Hearth goes almost directly from HTTP.sys to HttpRuntime.ProcessRequest. Fewer modules = fewer CPU cycles per request.
  2. No Access Logging: Stock IIS writes W3C request logs to disk by default, introducing noticeable I/O overhead at ~13K req/s. The portable server logs nothing.
  3. No Process Management Overhead: IIS constantly manages app-pool health pings, idle/CPU monitoring, and recycling checks. The portable worker focuses solely on serving requests.

What the raw numbers leave out (Where IIS earns its keep)

Throughput is only one piece of the puzzle. IIS's extra overhead funds enterprise features that the portable server simply lacks:

  • Kernel-Mode Output Caching: For cacheable responses, HTTP.sys can serve hundreds of thousands of requests per second, far bypassing these benchmarks. Because this test featured a dynamic, uncacheable page, this advantage wasn't utilized.
  • Resilience & Scalability: Web gardens (multiple worker processes), automatic crash recovery, app-pool recycling, and overlapped recycling.
  • Security: Native request filtering, TLS/SNI management, Windows Authentication, and IP restrictions.
  • Operations: Detailed logging, performance counters, Failed-Request Tracing, and dynamic compression.

Test Caveat: Both client and server shared the same 8 logical CPUs, meaning both ceilings are artificially constrained. Running this test using a separate machine over a physical network would raise the performance ceiling for both—and could either narrow or widen the gap.

Final Verdict

Out of the box, the Hearth Portable Server is leaner and edges out a stock, single-worker IIS by ~16% on this dynamic workload. If you stripped IIS down (disabled logging, removed unused modules), it would likely close most of this gap. Conversely, if kernel output caching were enabled on a cacheable page, IIS would pull far ahead.

The most reassuring takeaway isn't the 16% speed advantage—it's that both reports show an identical traffic shape. Both saturate around concurrency 8–16, peak at 128, dip at 256, and stay completely error-free.

Hearth doesn't just run fast; it behaves like a mature server under load, demonstrating graceful saturation and textbook Little's Law latency growth.

The Breakdown Performance Per Minute of Hearth Portable ASP.NET Web Server

Concurrency (Simultaneous Users) Performance Per Minute (Req/Min) Health Status
1 200,746 🟢 Perfect (No line)
2 365,293 🟢 Perfect
4 590,791 🟢 Efficiency multiplying
8 807,624 🟡 CPU is working hard
16 869,608 🟡 Approaching CPU limit
32 882,001 🔴 CPU is Maxed Out (Throughput flattens)
64 830,951 🔴 CPU is maxed (Small dip due to thread switching)
128 (PEAK) 898,796 🔥 Absolute Maximum Performance Ceiling
256 761,756 📦 Saturated (Safe queueing active)

Clone this wiki locally