[Performance]: there are spikes in time while building

[SimaTian]

Issue Description

While doing benchmarking for MSBuild, I've noticed a repeated spikes in time consumed.
This is annoying for several reasons:

• it is introducing noise to the benchmarking efforts (the spike is something like +50% time when it occurs)
• since it is occuring regularly on a cold build it is something that could be affecting pipelines and the spike is large if that is the case.
• it could be some error manifesting somewhere, but since it is random, it is hard to reproduce.

Steps to Reproduce

Run MSBuild cold build in a loop for a while.

Data

The times look like this:
95.2590266
53.3364027
93.6623841
53.8912314
51.5191799
49.8388009
49.926976
76.3744426

Or like this:
57.0661822
47.0323233
46.5913615
44.9238568
43.7712883
43.0863718
50.6900063
68.8321462
57.0158222
43.7932883
43.5368408
44.102942
43.3448263

I would chalk the first one to be the PC doing some weird stuff after a restart, but there is a secondary spike soon after and then one every five or so iterations(sometimes less, sometimes more, but the spike occurs semi regularly.)
Why is this happening?

Analysis

No response

Versions & Configurations

No response

Regression

• yes
• no

Regression Details

No response

[YuliiaKovalova]

Could you please add the exact steps to repro it:

• what msbuild instance (msbuild.exe or core)?
• MSBuild version;
• The built project;
• Environment characteristics.

Maybe it will be better measured in scope of PerfStar to understand if there is anything fishy with that. Not sure if it's actionable now.

[yuehuang010]

Are you using /M (multi-core)? This might cause some scheduling randomness.

[maridematte]

This is currently true, but not really actionable on our side.