Last week we delivered benchmarks comparing the performance of Ubuntu 9.04 vs. Fedora 11 and found for the most part that these two incredibly popular Linux distributions had performed about the same, except for a few areas where there notable differences. However, like in the past when we have looked at Ubuntu 7.04 to 8.10 benchmarks or benchmarking the past five Linux kernels, we are now looking at the performance of Fedora over their past few releases. In this article we have a range of system benchmarks from Fedora 9, 10, 11, and the latest Rawhide packages as of this week.
With all my respect to Phoronix, I believe you should explain to people why you are getting these results. In Linux what really matters is 1) kernel version 2) X.org version 3) X.org drivers versions 4) filesystems being tested.
What's more, updated Fedora may run slower or faster than its initial release, because during update cycle Fedora often pushes new kernel releases (2.6.x -> 2.6.x+1) and new X.org drivers.
So, in my mind, you should concentrate on finding regressions in kernel and X.org drivers and report them to appropriate bugzilla's instead of publishing these semi-useful "benchmarks" (which indeed explain nothing about system performance). Some of your data is really interesting and can help make Linux kernel the best kernel in the world.
Thanks for publishing this. Well presented. It is generous that you make the testing platform available for anyone to use.
I am sympathetic to comment from cruiseoveride, as I am finding I have been unsuccessfull to upgrade to FC11 from FC10. If I do a new install and get the full benefits of the new ext4 file system, you have to custom install your own applications, restore the data, which never all resides in /home, eg cacti or nagios, and you have to reconfigure config files from scratch such as sendmail, mysql, remote loging, etc., and any scripts you wrote to customize your use of F10. I'm not ready to move to Ubuntu, but I am going to wait for F12. I call F11 the vista release of Fedora -- you can get it to work as new install, but very frustrating to upgrade from previous release and then once installed the desktop has all kinds of quirks, including Administraor permissions challenges as I found no obvious way to login desktop as root. 0.02
Im a long time redhat/fedora user on my workstations and commercially (well, centos) and on my work servers. While the speed differences in this article are a little interesting, using a software 3d stack makes the graphic benchmarks entirely useless for real world comparison. It would be great if you could put a good nvidia graphics card in the machine, and benchmark with both the built in driver, and with nvidias proprietry drivers, probably from the livna repository. Sure this opens a can of worms with what your actually testing, but anyone who knows the distro well enough and is serious enough would be running like this, and anyone who reads the article to see what the system is capable of will get a better idea too.
BTW To the other commenters here, its worthwhile to note that fedora is a testing ground full of very raw development software, and its primary purpose is to push the development faster and better of bleeding edge linux. I find the sweet spot for me is to sit 1 fedora version behind current, but if you want a stable well supported system and do not want to do too much bleeding on the bleeding edge dont run fedora-latest. If your a developer or a hard core linux geek though, maybe its for you?