Benchmarks for dotnet/runtime#132544 (for @EgorBo)

[EgorBot]

Processing benchmark request from PR dotnet/runtime#132544.

Benchmark: OrchardCore CMS (requests/sec)
Targets: ubuntu24_aws_graviton4
Commits: main;PR_132544

• ubuntu24_aws_graviton4: live logs

Results will be posted here once they are ready.

[EgorBot]

Results for ubuntu24_aws_graviton4

@EgorBo

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

Runtime	RPS	StdDev	Noise (CV)	Min .. Max	Ratio	Median latency (p50 / p90 / p99)
main	10,692	51	0.5%	10,601 .. 10,731	baseline	4.88 ms / 6.98 ms / 18.26 ms
PR #132544	10,731	256	2.4%	10,462 .. 10,985	1.004 (+0.4%)	4.87 ms / 7.03 ms / 18.52 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
• 8 core(s) visible — app pinned to 0,1,2,3,4,5,6, bombardier pinned to 7
• 56 connections, 90s warmup, 2 process(es) x 3 x 15s measured
• DOTNET_GCDynamicAdaptationMode=0 (DATAS off), DOTNET_HillClimbing_Disable=1, DOTNET_ThreadPool_UnfairSemaphoreSpinLimit=0

Runtime	Process	Interval	RPS	p50	p90	p99	non-2xx
main	1	1	10,665	4.91	6.99	17.95	0
main	1	2	10,731	4.89	6.94	17.98	0
main	1	3	10,728	4.87	6.99	18.24	0
main	2	1	10,601	4.92	7.08	18.63	0
main	2	2	10,706	4.87	6.98	18.45	0
main	2	3	10,722	4.88	6.97	18.28	0
PR #132544	1	1	10,921	4.81	6.75	18.61	0
PR #132544	1	2	10,985	4.77	6.77	18.01	0
PR #132544	1	3	10,985	4.78	6.72	18.44	0
PR #132544	2	1	10,462	4.96	7.34	18.61	0
PR #132544	2	2	10,514	4.95	7.29	18.62	0
PR #132544	2	3	10,521	4.93	7.31	18.24	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

[EgorBo]

@EgorBot orchard -aws_graviton4 -profiler

[EgorBot]

Benchmark job submitted. The results will be posted here once they are ready.

• ubuntu24_aws_graviton4: live logs

[EgorBot]

Results for ubuntu24_aws_graviton4

@EgorBo

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

Runtime	RPS	StdDev	Noise (CV)	Min .. Max	Ratio	Median latency (p50 / p90 / p99)
main	11,293	70	0.6%	11,192 .. 11,371	baseline	4.66 ms / 6.53 ms / 17.56 ms
PR #132544	11,349	109	1.0%	11,211 .. 11,470	1.005 (+0.5%)	4.63 ms / 6.52 ms / 17.71 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
• 8 core(s) visible — app pinned to 0,1,2,3,4,5,6, bombardier pinned to 7
• 56 connections, 90s warmup, 2 process(es) x 3 x 15s measured
• DOTNET_GCDynamicAdaptationMode=0 (DATAS off), DOTNET_HillClimbing_Disable=1, DOTNET_ThreadPool_UnfairSemaphoreSpinLimit=0

Runtime	Process	Interval	RPS	p50	p90	p99	non-2xx
main	1	1	11,192	4.69	6.64	17.90	0
main	1	2	11,264	4.66	6.54	17.88	0
main	1	3	11,252	4.66	6.59	17.39	0
main	2	1	11,316	4.66	6.52	17.25	0
main	2	2	11,363	4.64	6.42	17.73	0
main	2	3	11,371	4.62	6.51	17.39	0
PR #132544	1	1	11,417	4.61	6.47	17.90	0
PR #132544	1	2	11,470	4.59	6.40	17.01	0
PR #132544	1	3	11,451	4.58	6.43	17.65	0
PR #132544	2	1	11,211	4.69	6.58	17.67	0
PR #132544	2	2	11,282	4.65	6.56	17.81	0
PR #132544	2	3	11,266	4.66	6.58	17.75	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.

Profiling artifacts

Supported perf events on this machine: perf_events.txt

OrchardCore.Cms:

	main	PR_132544
flamegraph	link	link
functions	link	link
asm	link	link
stats	link	link
speedscope	link	link

Full logs