Benchmarks for dotnet/runtime#132945 (for @EgorBo)

[EgorBot]

Processing benchmark request from PR dotnet/runtime#132945.

Benchmark: OrchardCore CMS (requests/sec)
Targets: ubuntu24_azure_turin, ubuntu24_aws_graviton5
Commits: main;PR_132945

• ubuntu24_azure_turin: live logs
• ubuntu24_aws_graviton5: live logs

Results will be posted here once they are ready.

[EgorBot]

Results for ubuntu24_aws_graviton5

@EgorBo

OrchardCore CMS — throughput (requests/sec, higher is better)

Runtime	RPS	StdDev	Noise (CV)	Min .. Max	Ratio	Median latency (p50 / p90 / p99)
main	34,994	167	0.5%	34,775 .. 35,251	baseline	1.56 ms / 2.71 ms / 4.63 ms
PR #132945	34,791	278	0.8%	34,474 .. 35,099	0.994 (-0.6%)	1.57 ms / 2.73 ms / 4.59 ms

RPS is the mean of 6 measured intervals per runtime; Noise (CV) is their coefficient of variation — treat differences smaller than the noise as inconclusive.

Configuration and per-interval results

• OrchardCore a71838a0, Blog recipe, SQLite, /about, linux-arm64, net11.0, self-contained
• 16 core(s) visible — app pinned to 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14, bombardier pinned to 15
• 60 connections, 90s warmup, 2 process(es) x 3 x 15s measured
• DOTNET_HillClimbing_Disable=1, DOTNET_gcServer=1

Runtime	Process	Interval	RPS	p50	p90	p99	non-2xx
main	1	1	34,775	1.57	2.72	4.73	0
main	1	2	34,955	1.55	2.71	4.70	0
main	1	3	34,929	1.56	2.71	4.61	0
main	2	1	34,935	1.56	2.70	4.63	0
main	2	2	35,121	1.56	2.69	4.62	0
main	2	3	35,251	1.54	2.69	4.58	0
PR #132945	1	1	34,474	1.58	2.75	4.60	0
PR #132945	1	2	34,671	1.57	2.74	4.65	0
PR #132945	1	3	34,504	1.58	2.75	4.65	0
PR #132945	2	1	34,937	1.56	2.71	4.56	0
PR #132945	2	2	35,059	1.56	2.71	4.58	0
PR #132945	2	3	35,099	1.56	2.70	4.52	0

The runtime assemblies are taken from each commit's Core_Root, which is not ReadyToRun-compiled — absolute numbers are therefore lower than a released runtime; only the comparison between the rows above is meaningful.

Full logs

[EgorBot]

Results for ubuntu24_azure_turin

@EgorBo

OrchardCore CMS — throughput (requests/sec, higher is better)

Runtime	RPS	StdDev	Noise (CV)	Min .. Max	Ratio	Median latency (p50 / p90 / p99)
main	27,371	460	1.7%	26,442 .. 27,655	baseline	1.88 ms / 3.69 ms / 6.39 ms
PR #132945	27,799	186	0.7%	27,539 .. 28,055	1.016 (+1.6%)	1.87 ms / 3.65 ms / 6.33 ms

RPS is the mean of 6 measured intervals per runtime; Noise (CV) is their coefficient of variation — treat differences smaller than the noise as inconclusive.

Configuration and per-interval results

• OrchardCore a71838a0, Blog recipe, SQLite, /about, linux-x64, net11.0, self-contained
• 16 core(s) visible — app pinned to 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14, bombardier pinned to 15
• 60 connections, 90s warmup, 2 process(es) x 3 x 15s measured
• DOTNET_HillClimbing_Disable=1, DOTNET_gcServer=1

Runtime	Process	Interval	RPS	p50	p90	p99	non-2xx
main	1	1	27,450	1.88	3.70	6.39	0
main	1	2	27,574	1.87	3.70	6.39	0
main	1	3	27,535	1.88	3.68	6.39	0
main	2	1	27,569	1.88	3.68	6.38	0
main	2	2	27,655	1.86	3.68	6.47	0
main	2	3	26,442	1.88	3.78	8.02	0
PR #132945	1	1	27,704	1.87	3.65	6.33	0
PR #132945	1	2	27,539	1.87	3.69	6.49	0
PR #132945	1	3	27,743	1.87	3.65	6.33	0
PR #132945	2	1	27,785	1.86	3.63	6.37	0
PR #132945	2	2	27,965	1.85	3.64	6.29	0
PR #132945	2	3	28,055	1.85	3.62	6.20	0

The runtime assemblies are taken from each commit's Core_Root, which is not ReadyToRun-compiled — absolute numbers are therefore lower than a released runtime; only the comparison between the rows above is meaningful.

Full logs