Using W10Privacy To Boost Ubuntu WSL Performance On Windows 10

[phoronix]

Phoronix: Using W10Privacy To Boost Ubuntu WSL Performance On Windows 10

While Microsoft is working on low-level improvement to the Windows Subsystem for Linux (WSL) to yield better I/O performance, it is possible to dramatically increase some disk workloads by removing a number of running Windows services via the independent W10Privacy application. Here are some benchmarks of W10Privacy on the overall performance impact to Microsoft Windows 10 Pro x64 itself and of Ubuntu 18.04 running on the Windows 10 installation via WSL.

http://www.phoronix.com/vr.php?view=26466

[nanonyme]

Any chance of getting a summary of what was actually disabled in the scope of the test?

[ermo]

Any chance of getting a summary of what was actually disabled in the scope of the test?

Not only that, it would be interesting to hear an official MS-employed Windows developer's honest engineering take on the potential impact.

[AsuMagic]

Can we take just a minute to appreciate how pathetic this is?

[mczak]

From the results, looks to me like some index services crap (not that this would be a windows-exclusive problem...).

[MadCatX]

The IO performance penalty on default Win10 config is quite staggering. I honestly thought that the days where trimming the fat on Windows was a real thing were long gone. It's too bad that Michael didn't specify which services exactly were disabled by that privacy doohickey. I suppose it could be interesting to re-run the test on a stock Win10 config with just the Windows Defender disabled.

[Michael]

The IO performance penalty on default Win10 config is quite staggering. I honestly thought that the days where trimming the fat on Windows was a real thing were long gone. It's too bad that Michael didn't specify which services exactly were disabled by that privacy doohickey. I suppose it could be interesting to re-run the test on a stock Win10 config with just the Windows Defender disabled.

The screenshot encompasses many of them, but anyone opening up the program can basically reproduce by just launching it and going through and clicking the stuff.... I disabled most stuff except for the obvious irrelevant things like 'disable web camera access' and 'uninstall Skype'.

As mentioned in the article and also linked to the OB file in there, I have done Windows Defender comparison tests before to not much of a difference in speed.

[sireangelus]

Ehi michael, why dont you test wsl on windows server? it would be more of an apple to apple comparison than testing windows 10

[luckz]

"Changed some options, except I didn't press some other buttons" isn't that helpful, especially from you as a champion of reproducible testing.

The easiest way to derive knowledge here is to solely test the most dramatically affected single test with different configurations to see what actually triggers this enormous performance increase.

I would also like to point out that PowerShell scripts like https://github.com/Disassembler0/Win...l-Setup-Script , https://github.com/W4RH4WK/Debloat-Windows-10 , https://gist.github.com/alirobe/7f3b34ad89a159e6daa1 (just some of many) are very customizable and again in the spirit of reproducibility.

I hope once the 'culprits' of reduced WSL (and general Win10 I/O) performance are found, you will do us the service of listing Win 10 as well as tweaked/debloated Win 10 in future benchmarks. Thanks.

[starshipeleven]

Can we take just a minute to appreciate how pathetic this is?

Seems mostly IO, so it's indexing or some shit. Not new, not windows-dependent.