Workstation Benchmarks: Windows 7 vs. Ubuntu Linux

[Love4Boobies]

I'm afraid inserting new entries is the same as writing a file containing multiple blocks (clusters)... Since clusters are contiguous by nature, the ratio would stay exactly the same.

Anyway, I've had Windows machines running for months on end (and even then I restarted for updates) without ever failing. That's a whole lot more than I can say about Linux distros where X fails constantly. And everytime I update something I have huge problems with drivers (one driver needs one update, the other doesn't work with it and I need to downgrade again - if I'm lucky and I can find a driver at all).

[smitty3268]

I'm afraid inserting new entries is the same as writing a file containing multiple blocks (clusters)... Since clusters are contiguous by nature, the ratio would stay exactly the same.

I'm not sure I follow you - are you claiming that writing 1 million 1KB files is the same as writing a single 1GB file? Because it's definitely not, there's tons of extra overhead that goes on per-file.

Anyway, I've had Windows machines running for months on end (and even then I restarted for updates) without ever failing. That's a whole lot more than I can say about Linux distros where X fails constantly. And everytime I update something I have huge problems with drivers (one driver needs one update, the other doesn't work with it and I need to downgrade again - if I'm lucky and I can find a driver at all).

I agree that Windows is pretty stable these days - it depends heavily on the hardware drivers you're using, of course. But you have to reboot once a month for all the security updates. Linux can sometimes last me 2 or 3 months between reboots. Anyway, this is getting pretty off-topic and moving into a "my OS is better than yours" flamewar territory.

[marwi509]

Hi.

I just registered to say that the article is dead wrong. IOzone needs Cygwin under Windows (it uses mmap() and pthreads), which is an emulation layer. Benchmarks should never be done with emulation of any sort.

Try running a file system benchmarking software that's native to Windows. Use Wine under Linux and see what happens...

Cheers,
Bogdan

Cygwin isn't an emulator though, it's a compatibility layer.

[FunkyRider]

Oh great! I'm a regular UGS, Pro/E user, I see it performs a lot better in Ubuntu than in Windows. Can I have a copy of Pro/E for Ubuntu? I wanna convert!

[deanjo]

Oh great! I'm a regular UGS, Pro/E user, I see it performs a lot better in Ubuntu than in Windows. Can I have a copy of Pro/E for Ubuntu? I wanna convert!

They don't make Pro/E for linux any more do to lack of demand. Windows only nowdays. Wildfire 2 IIRC was the last version that had linux support and is no longer sold.

[cruiseoveride]

I think there might be something wrong with the openssl rsa-4096bit test. Isnt it a very cpu intensive task? How would the results vary so much between operating systems?

[deanjo]

I think there might be something wrong with the openssl rsa-4096bit test. Isnt it a very cpu intensive task? How would the results vary so much between operating systems?

The reason is the build openSSL micheal is using in PTS.

Win32OpenSSL_Light-1_0_0a.exe which is the 32-bit version built virtually no optimizations.

For a fair comparison PTS should have used the 64-build which does carry SSE optimizations and such.

http://www.shininglightpro.com/download/Win64OpenSSL_Light-1_0_0a.exe

[RealNC]

Emulation issue: Cygwin is not really an emulator. It's an implementation of common Unix/Linux APIs. It will affect performance though in cases like mmap and other kernel calls that don't exist in Windows. So to benchmark, it would be better to use a native Windows disk benchmark. Even if the programs used are different for each OS, the results would be more suitable for comparison if they provide tests common in both programs.

Linux I/O blocking issue: This doesn't happen with all systems. I've no idea why. But when it happens, it sucks major. And it's not an issue of "doing something wrong." What is there to do wrong? Should I do the monkey dance before copying a file to get it right? And in any event, the problem has been reported since ages and hopefully the lkml post mentioned previously means this is about to get fixed at last.

[jacob]

So the test has issues on OS X as well.
In a CPU raw test there's no way the OS can have that effect, as simple as that...

It can, if it's due to differences between compilers.

[yotambien]

I don't know the technicalities, but yes, certain things under Cygwin are surely a lot slower than their native counterparts in Linux. Copying stuff over SSH in my home network is a pain when a Windows machine is involved, either as a server or as a client.