Arch Linux 2009.08 Benchmarks

[Apopas]

Cool would be to compare two rolling distros. Arch vs gentoo for example

[kraftman]

It's nice to see such benchmark

[MuPuF]

The goal of the distributions's benchmark is to test default choices of made by the distributions. They are all based on the same projects.

What is great with testing Arch is that it is guaranteed to be as vanilla as possible. It is also using the latest stable versions.

Thanks Michael for doing the test, hope you'll keep this version so as we may see more tests involving ArchLinux

[RealNC]

This is a nice article for people wanting to switch to Arch due to its rolling release nature and were afraid that it might be slower than Ubuntu.

[Vadi]

Well, it's nice. Not that big of a performance difference to switch to a less pampered distro though.

[Arch64]

Hmm, even though I am an Arch user, I can see that this benchmark is a bit unfair for Ubuntu, since it uses an older fs (ext3) and an older kernel.

It would be better to have made an additional benchmark with an ext4 Ubuntu installation or/and an Ubuntu 9.10 installation (note: Ubuntu 9.10 has though testing packages, like GNOME 2.28 beta which isn't even in [testing] in Arch yet).

[SyXbiT]

Thanks for the benchmarks
One this is benchmarking different kernels or file systems.
but this is doing everything at the same time.
distro, kernel, gcc, file system etc..

I would like to see Arch compared to a distro with very similar versions of kernel and with ext4.

Could I recommend a benchmark of Ubuntu 9.10 to the current Arch Linux at the time.
My guess is that around then they'll both be on 2.6.31, ext4 and gcc 4.4

Right now there are too many variables

[Apopas]

The benchmark would be better if instead of Ubuntu we had Fedora which uses newer kernel and ext4.

[sharms]

I think a big problem with Arch is (that most people gloss over) is the lack of package authenticity.

You have no good automated way via pacman to determine if you are getting the package from Arch, or if from some other party. If someone hacks a mirror, you have no way of knowing.

[Zhick]

The difference in the sqlite-benchmark is easily explained: kernels >= 2.6.30 default to data=writeback for ext3, but not for ext4. I'm sure if you'd configure Archs ext4-fs with data=writeback as well it would probably end up at least as fast.