Benchmarks

Requirements

• dotnet >= 10.0
• go >= 1.25
• PowerShell 7
• bombardier >= 1.2

View the latest results.

Run the benchmark script with PowerShell 7, with bombardier, dotnet, and go on PATH (only the runtimes needed by the selected apps are required):

.\scripts\RunBenchmarks.ps1

For example, to benchmark specific apps and connection counts:

.\scripts\RunBenchmarks.ps1 -Apps AspNetMvcBench,AspNetRazorPages,FalcoBench,FrankBench,GiraffeBench,GiraffeEndpointsBench,GoBench,OxpeckerBench,RazorSlicesBench -Connections 64,256 -Runs 5

AspNetMvcBench benchmarks ASP.NET Core MVC, and AspNetRazorPages benchmarks ASP.NET Core Razor Pages. Like the other apps, both serve a “Hello World!” page with the current UTC time at /. Their publish trimming is disabled because ASP.NET Core MVC and Razor Pages do not support trimming.

The Giraffe apps render their templates with Giraffe.ViewEngine, Oxpecker uses Oxpecker.ViewEngine, and Frank uses Falco.Markup.

Results are saved under results\<timestamp>\. The script prints the full paths to summary.md and summary.json; each measured run also has a raw <app>-c<connections>-run<run>.json file in the same directory. The Markdown summary groups results by connection count and separates throughput and latency from resources and response counts for readability.

For a quick, non-persistent run, use -Mini. It runs 64 connections for one 30-second measurement after a one-second warmup, skips the idle wait, and prints the summary without creating files under results\.

To verify that the selected apps can be built/published without starting them or running a load test, use -BuildOnly. This does not require bombardier and does not create benchmark result files:

.\scripts\RunBenchmarks.ps1 -BuildOnly

The script rebuilds selected apps under build\. It runs the load generator locally, so the results also reflect contention for local resources.

Results

• Date: 2026-11-12 13:41:42Z
• OS: Microsoft Windows 10.0.19045
• CPU cores: 4
• .NET SDK: 10.0.115
• Go: go version go1.26.0 windows/amd64
• Duration: 30s x 3 runs (median reported), warmup 10s, idle 30s

64 connections

Throughput and latency

App	Req/s median	min	max	p50 ms	p99 ms
FalcoBench	81013	80731	81761	1	1.99
GiraffeEndpointsBench	80046	79274	81714	1	1.99
GiraffeBench	79509	76717	80828	1	1.99
OxpeckerBench	79169	78768	80053	1	1.99
RazorSlicesBench	78115	77790	79700	1	1.99
FrankBench	77982	69833	80146	1	2
AspNetRazorPagesBench	48785	48601	49265	1	2
AspNetMvcBench	48719	48464	49220	1	2
GoBench	48287	48098	49172	1.16	3.74

Resources and responses

App	Peak MB	Idle MB	Startup ms	Publish MB	Non-200
GoBench	17.4	15.6	1378	8.2	0
GiraffeBench	48.5	39.6	1235	13.1	0
FalcoBench	49.3	40.3	1277	13.1	0
GiraffeEndpointsBench	49.5	40	1274	13.1	0
RazorSlicesBench	49.9	41.1	1298	13	0
OxpeckerBench	50.2	41.5	1266	13.1	0
FrankBench	51.4	45	1323	13.6	0
AspNetRazorPagesBench	173.9	157	763	48	0
AspNetMvcBench	176.7	168.7	741	48	0

256 connections

Throughput and latency

App	Req/s median	min	max	p50 ms	p99 ms
FalcoBench	64860	64764	65123	3.99	6.98
OxpeckerBench	64809	59423	65597	3.99	6.98
FrankBench	64759	63863	65715	3.99	6.98
RazorSlicesBench	64575	64324	65402	3.99	5.98
GiraffeBench	64363	63555	65498	3.99	6.98
GiraffeEndpointsBench	64073	63950	65835	3.99	6.98
GoBench	46609	46608	47203	4.74	14.96
AspNetRazorPagesBench	44359	43985	45382	4.99	8.98
AspNetMvcBench	42176	31492	42378	4.99	11.97

Resources and responses

App	Peak MB	Idle MB	Startup ms	Publish MB	Non-200
GoBench	24.8	23.8	519	8.2	0
OxpeckerBench	70.6	48.2	1236	13.1	0
FalcoBench	70.8	44.5	1195	13.1	0
FrankBench	70.9	47.2	1262	13.6	0
GiraffeEndpointsBench	73.7	44.8	1184	13.1	0
RazorSlicesBench	75.5	45.8	1220	13	0
GiraffeBench	76	44.3	1161	13.1	0
AspNetRazorPagesBench	187.4	165.7	624	48	0
AspNetMvcBench	191.3	164.7	637	48	0