Performance difference: ApiCrudGetProductDetails, ApiCrudAddProduct, ApiCrudUpdateProduct, ApiCrudDeleteProduct

[pr-benchmarks]

A performance regression has been detected for the following scenarios:

Scenario	Environment	Date	Old RPS	New RPS	Change	Deviation	StDev
ApiCrudAddProduct	API add new element aspnet-citrine-lin	01/24/2021 01:11:09	328,865	361,254	9.85 % (32,390) 👍	11.47 σ	2,822.85
ApiCrudDeleteProduct	API remove existing element aspnet-citrine-lin	01/24/2021 01:12:56	359,698	409,799	13.93 % (50,101) 👍	22.85 σ	2,192.47
ApiCrudGetProductDetails	API return element details aspnet-citrine-lin	01/24/2021 01:10:17	343,676	386,351	12.42 % (42,675) 👍	14.87 σ	2,870.32
ApiCrudUpdateProduct	API update existing element aspnet-citrine-lin	01/24/2021 01:12:02	332,444	375,623	12.99 % (43,179) 👍	22.99 σ	1,878.43

Changes:

Runtime	Previous Version	Current Version	Diff
ASP.NET	n/a	n/a	n/a
CLR	n/a	n/a	n/a
SDK	6.0.100-alpha.1.21069.6	6.0.100-preview.1.21073.4	n/a

@javiercn
@pranavkm

[javiercn]

@sebastienros is this the result of a fix where we regressed things previously or did we make an improvement somewhere

[sebastienros]

There is definitely a perf improvement on MVC scenarios by looking at the charts.

• it only affects the MVC specific app, not the Json MVC scenario from the main Benchmarks app
• it only affects Linux
• it is due to a change in the runtime from this range: dotnet/runtime@16e6e95...6cf1b8e

I will assume it's the change on diagnostics id generation from @tarekgh (dotnet/runtime@035821a) since the app that is not affected has logging providers cleared, and not the MVC app.

Which means the improvement might be more global, it's just visible on this app since it doesn't have logging disabled... So I checked "the other app" we know had an issue with logging, and that was the origin of this issue that was worked on, aka YARP. And we can see an improvement there too.

/cc @alnikola @halter73