Windows 10 WSL: Ubuntu vs. openSUSE Leap Performance

[roelandjansen]

Seemed like most everything except the postgresql stuff was faster. Am I reading this stuff wrong?

you are not. I did not see much of a difference either. Besides. it's WSL.. the only reason to use this kind of stuff is not really to do workloads on but to be able to do some stuff in a better way compared to the windows tooling, towards linux/unix systems... and only, and only IF virtualization is not a possibility.

[Michael]

What about native Windows binaries vs Linux under WSL tests? Will there be any?

No as don't have those test profiles ported over to Windows natively unless some company steps up with a commercial agreement / sponsorship to improve the PTS windows support.

[m132]

What about native Windows binaries vs Linux under WSL tests? Will there be any?

[clintar]

Wow... Why is OpenSuSE so bad? Does this match up with known performance discrepencies between those GCC revisions?

To the extent that it's affected by Skylake-X weirdness, I wonder if any scheduler improvements made it into the 42.3 kernel that will help.

Seemed like most everything except the postgresql stuff was faster. Am I reading this stuff wrong?

[polarathene]

Arch has a WSL support I think? Would be nice if it gets official support on the store too, see how a rolling distro compares

Curious if Clear Linux would also have an advantage in WSL like it did with your Docker tests, no idea if this has any WSL support though.

I don't know much about WSL, the I/O comments were indicating Linux FS was writing on top of Windows FS like a VM image file? Is it possible to assign a raw partition/disk or 10Gbps ethernet network share to get better performance? If network share is an option, you could also run Windows as a para-virtualized machine with an internal network(connect to host network share from guest), with network setup right using virtio and the like that'd perform pretty well.

[coder]

Wow... Why is OpenSuSE so bad? Does this match up with known performance discrepencies between those GCC revisions?

To the extent that it's affected by Skylake-X weirdness, I wonder if any scheduler improvements made it into the 42.3 kernel that will help.

[Michael]

When I compile something in WSL I noticed that it is very slow and MSMPEng.exe was using multiple CPU cores. This is even more noticeable with JavaScript apps due do huge npm tasks where tons of small files are involved.

Can you compare your benchmarks with
1. adding WSL's filesystem into Windows Defender's exclusion list
2. completely disabling Windows Defender

I would guess that WSL's performance will stay crippled for these tasks until their antivirus team starts improving its on-demand scanning performance (or just start to ignore WSL's pico processes).

When I compile something in WSL I noticed that it is very slow and MSMPEng.exe was using multiple CPU cores. This is even more noticeable with JavaScript apps due do huge npm tasks where tons of small files are involved.

Can you compare your benchmarks with
1. adding WSL's filesystem into Windows Defender's exclusion list
2. completely disabling Windows Defender

I would guess that WSL's performance will stay crippled for these tasks until their antivirus team starts improving its on-demand scanning performance (or just start to ignore WSL's pico processes).

That Windows install is no longer around but in weeks/months when next doing a Windows installation can do such a comparison if I remember.

[hsiktas]

When I compile something in WSL I noticed that it is very slow and MSMPEng.exe was using multiple CPU cores. This is even more noticeable with JavaScript apps due do huge npm tasks where tons of small files are involved.

Can you compare your benchmarks with
1. adding WSL's filesystem into Windows Defender's exclusion list
2. completely disabling Windows Defender

I would guess that WSL's performance will stay crippled for these tasks until their antivirus team starts improving its on-demand scanning performance (or just start to ignore WSL's pico processes).

[phoronix]

Phoronix: Windows 10 WSL: Ubuntu vs. openSUSE Leap Performance

Following my recent Windows 10 WSL CPU scaling benchmarks to see how well Microsoft's Windows Subsystem for Linux for running native Linux binaries compared to various distributions, I did a comparison of Ubuntu and openSUSE on WSL compared to running these distributions bare metal.

http://www.phoronix.com/vr.php?view=25011