This sounds good. I hope you get time to publish some results! The reason for running it regularly is more as an approach to automated regression testing. It does taka time to make test suites so if we cuuld use the Phorionix suite as a reg.test suite then that would be nice. Can we pick parts from it or adapt it to run in a shorter time span with that in mind? cu 2009/11/8 M. Edward (Ed) Borasky :

On Sun, Nov 8, 2009 at 5:51 AM, Terje J. Hanssen wrote:

...

Birger Kollstrand wrote:

...

Hi,

Please see:

http://www.phoronix.com/scan.php?page=article&item=centos_54_comparison&num=1

I think it would be nice to run the Phoronix suite at each new build. Does any one have any experience in that here?

btw. centOS comes out best in 11 ot of 18 tests.

I haven't any experience with this. But I'm wondering why openSUSE achieves relative bad results on the central Apache benchmark (which I think I've seen also in earlier tests)? http://www.phoronix.com/scan.php?page=article&item=centos_54_comparison&num=2

Terje

I have extensive Linux benchmarking experience, and I've spent a fair amount of time over the past couple of weeks running parts of the Phoronix suite on my machine here at home. I'm about to publish some test results comparing the four supported journaled filesystems (xfs, reiserfs, ext3 and ext4) on the Phoronix "disk" and "filesystem" suites.

I'm not sure how much free time I'll have to devote to this - I need to spend some more time out beating the bushes for *paid* work. But within those constraints, if there's hardware available, I can look into some of these issues.

Incidentally, the two suites I am running - "disk" and "filesystem" - take about 5.5 hours to complete on my machine (a single Seagate Barracuda with a dual-core Athlon64 X2 2.2 GHz booted down to 1 core and 1 GB of RAM. I haven't even attempted to run the whole enchilada, and I don't see any practical reason why you'd want to do so on a regular basis. But if there's hardware available, it's probably worth doing *once* with some of my profiling tools enabled.

-- M. Edward (Ed) Borasky http://borasky-research.net

"I've always regarded nature as the clothing of God." ~Alan Hovhaness

-- To unsubscribe, e-mail: opensuse-testing+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-testing+help@opensuse.org

Thx, I have started to look at it now. Th user interface looks very promising, both GUI and CLI..... but it takes ages to get it going. The download speed is awful........ 2009/11/8 M. Edward (Ed) Borasky :

On Sun, Nov 8, 2009 at 11:13 AM, Birger Kollstrand wrote:

...

This sounds good. I hope you get time to publish some results!

The reason for running it regularly is more as an approach to automated regression testing. It does taka time to make test suites so if we cuuld use the Phorionix suite as a reg.test suite then that would be nice.

Can we pick parts from it or adapt it to run in a shorter time span with that in mind?

Yeah, you can pick parts, and add your own benchmarks. They actually have a way to automate kernel performance regression testing already built!

http://www.phoronix.com/vr.php?view=14285

and

http://www.phoronix.com/vr.php?view=14347

But you need to do a lot of up-front planning / analysis / profiling to do this efficiently. For example, if there's a change in the Linux kernel block layer, you don't need to run all of the tests / suites to detect performance regressions, just those which have a block-layer bottleneck!

-- M. Edward (Ed) Borasky http://borasky-research.net

"I've always regarded nature as the clothing of God." ~Alan Hovhaness

-- To unsubscribe, e-mail: opensuse-testing+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-testing+help@opensuse.org