Workstation Benchmarks: Windows 7 vs. Ubuntu Linux

[smitty3268]

First of all, file systems are journaled, not OSes. Your observation that both of them are journaled is great, esp. since my quote says exactly that. Journaling can be enabled/disabled for both NTFS and ext4. I'm not sure why I should care that journaling is enabled by default. Even if it were normal to assume that journaling was turned on for both file systems then the benchmark is incomplete. Which one's faster with journaling disabled? Any benchmark should give the appropriate information.

Next time please at least look up the information you claim to understand.

Cheers,
Bogdan

Michael always uses the defaults in these benchmarks, and the default for both systems is to have journaling enabled. If you want to test how different tweaks affect performance, you should try it out yourself. Michael generally won't do those tests himself. If you think that makes the whole test worthless, then you probably shouldn't even be reading Phoronix, becuase it's just going to keep pissing you off.

[pvtcupcakes]

First of all, file systems are journaled, not OSes. Your observation that both of them are journaled is great, esp. since my quote says exactly that. Journaling can be enabled/disabled for both NTFS and ext4. I'm not sure why I should care that journaling is enabled by default. Even if it were normal to assume that journaling was turned on for both file systems then the benchmark is incomplete. Which one's faster with journaling disabled? Any benchmark should give the appropriate information.

Next time please at least look up the information you claim to understand.

Cheers,
Bogdan

Unless stated otherwise, Michael only tests the default configuration.

[Love4Boobies]

If you think that makes the whole test worthless, then you probably shouldn't even be reading Phoronix, becuase it's just going to keep pissing you off.

No, I think the test is worthless because it's wrong for the reasons I stated here.

[smitty3268]

No, I think the test is worthless because it's wrong for the reasons I stated here.

I agree the Cygwin thing is a good catch. IOZone sucks as a benchmark anyway, I'd much prefer a more realistic test. Still, the FS comparison was only a portion of the benchmarks here, and I don't think anyone should be very surprised to see NTFS slower than EXT4.

[frantaylor]

The disk I/O benchmarks would be much more interesting if you also reported CPU utilization.

[kraftman]

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...

Maybe it's accelerated on Linux?

To do a comparison that is meaningful you should always test native code... who cares if a test written for Linux and ported with some type of compatibility layer is slower on Windows!

Was this emulated on Windows?

[Love4Boobies]

Was this emulated on Windows?

Yep. The benchmarking software does not contain a Windows port, it needs Cygwin, thus making it considerably slower.

[kraftman]

Btw. do some SMP benchmarks, please. This should be very interesting to see.

[kraftman]

Yep. The benchmarking software does not contain a Windows port, it needs Cygwin, thus making it considerably slower.

Ok, it will be great to see native applications used.

[monraaf]

How about benchmarks of some other file system operations. Like creating and deleting complex directory structures with tens of thousands of files. I'd love to see how windows copes (or fails to cope) with that