Ubuntu 15.10 vs. Fedora 23 With The Intel Xeon E3 v5 Skylake

[phoronix]

Phoronix: Ubuntu 15.10 vs. Fedora 23 With The Intel Xeon E3 v5 Skylake

A few days ago I wrote about building an Intel Skylake Xeon E3 v5 "Skylake" system and my experiences under Ubuntu. Here's a few notes about this Xeon E3 1245 v5 system when trying Fedora 23 Linux, along with some comparative performance benchmarks...

http://www.phoronix.com/scan.php?page=news_item&px=Xeon-Skylake-Fedora-23

[fhuberts]

Those comparisons are not really fair: Fedora has SELinux turned on by default while Ubuntu has not.
This makes Fedora a lot more secure at the expensive of some performance.

For fair performance-only comparisons you should turn SELinux off by using the kernel command-line option 'selinux=0' (note: _never_ do this on real systems)

[liam]

Those comparisons are not really fair: Fedora has SELinux turned on by default while Ubuntu has not.
This makes Fedora a lot more secure at the expensive of some performance.

For fair performance-only comparisons you should turn SELinux off by using the kernel command-line option 'selinux=0' (note: _never_ do this on real systems)

Selinux doesn't have THAT much overhead. Besides, I believe Ubuntu uses apparmour which should have similar overhead.

[chilek]

Selinux doesn't have THAT much overhead. Besides, I believe Ubuntu uses apparmour which should have similar overhead.

Liam, but how many apps does AppArmor secure in Ubuntu? But of course both don't have so much overhead.

[sarmad]

Those comparisons are not really fair: Fedora has SELinux turned on by default while Ubuntu has not.
This makes Fedora a lot more secure at the expensive of some performance.

For fair performance-only comparisons you should turn SELinux off by using the kernel command-line option 'selinux=0' (note: _never_ do this on real systems)

Ubuntu has AppArmor. So no, SELinux shouldn't be turned off during the test, unless you want to also turn off AppArmor on Ubuntu. But then what's the point of comparing non-default setups?

[blackout23]

Those comparisons are not really fair: Fedora has SELinux turned on by default while Ubuntu has not.
This makes Fedora a lot more secure at the expensive of some performance.

For fair performance-only comparisons you should turn SELinux off by using the kernel command-line option 'selinux=0' (note: _never_ do this on real systems)

It's fair, because the benchmark shows exactly what Michael wants and that is how different distros compare out of the box. That's the only logical comparison you can make.
Why should benchmarks be deliberately tweaked just to make all distros perform equally? Do people really get their feelings hurt over benchmarks?

[Michael]

For curiosity sake, running some fresh benchmarks of Fedora on a slower system of SELinux on/off... last time I did such though, the results were rather boring.

[Michael]

Does SELinux Have Much Of A Performance Impact On Fedora 23? - http://www.phoronix.com/scan.php?page=news_item&px=Fedora-23-SELinux-Impact

[liam]

Liam, but how many apps does AppArmor secure in Ubuntu? But of course both don't have so much overhead.

No idea, but I'd HOPE they have confined all default apps, at a minimum

[Hugh]

As Michael said (more or less) what's needed is analysis. For example, the difference in the very last benchmark, Apache, is enormous. What's going on? With such a large difference, it ought to be easy to figure out. But no, I'm not volunteering.