OpenSUSE 42.1 Tweaked For Better Out-Of-The-Box Performance
Back in September I posted Fedora vs. openSUSE vs. Manjaro vs. Debian vs. Ubuntu vs. Mint Linux Benchmarks. Of that six-way Linux distribution comparison, several Phoronix readers complained that I was somehow anti-openSUSE or that testing out-of-the-box distribution performance isn't right, since openSUSE 42.1 Leap tended to lose the most in that testing. Well, thanks to those tests, the out-of-the-box performance for openSUSE 42.1 is now going to be better.
OpenSUSE 42.1 Leap is set to ship tomorrow and will offer better-configured defaults so that it's more competitive with other Linux distributions. Following that original article and the complaints over the openSUSE default performance, I followed up with Benchmarks Of The OpenSUSE Leap Kernel Flavors.
Richard Brown, the openSUSE Board Chairman, had then been in contact with me about the data. The data held its ground and they were able to figure out how to improve the out-of-the-box performance. Richard Brown explained in an email:
OpenSUSE 42.1 Leap is set to ship tomorrow and will offer better-configured defaults so that it's more competitive with other Linux distributions. Following that original article and the complaints over the openSUSE default performance, I followed up with Benchmarks Of The OpenSUSE Leap Kernel Flavors.
Richard Brown, the openSUSE Board Chairman, had then been in contact with me about the data. The data held its ground and they were able to figure out how to improve the out-of-the-box performance. Richard Brown explained in an email:
The biggest change since your benchmarks was the realisation that we were shipping RC1 with a udev rule that set the schedular for SSDs to NOOPSo the out-of-the-box performance should now be better; I'll have benchmarks of 42.1 Leap this week. Additionally, they're now looking at utilizing the Phoronix Test Suite for better tracking the out-of-the-box SUSE Linux performance moving forward. I've also offered some SUSE coverage on LinuxBenchmarking.com.
This was because of the theoretical improvement that NOOP should bring SSDs. The reality is somewhat different, and some SSDs are more negatively impacted by running in NOOP than others - the ones you used seemed to be some of the worst we analysed :)
So we changed the rule and now we're using Deadline for SSDs. On The SSDs that like NOOP its not as fast as it could be, but its a much nicer default
It was your benchmarks which triggered this investigation, and those additional ones you did provided really valuable context which helped us narrow down the problem very quickly, so thank you :)
I'm now looking at integrating PTS into openQA for both openSUSE and SLE development to help keep a closer eye on this sort of 'out of the box' performance in the future :)
OpenSUSE Enters The Beer Business With "Old Toad"
2 Comments