A Few Worrisome Regressions Appear In Ubuntu 15.04 vs. 15.10 Performance

[andre30correia]

> A Few Worrisome Regressions Appear In Ubuntu 15.04 vs. 15.10 Performance

Yeah, right. Ubuntu is always in broken state. I guess we can call it "stable" as well....

this regressions are most likely a kernel problem it will appear in arch when they switch to 4.2

[andre30correia]

I'd say the opposite is true. In my experience, Arch is always stable, while Ubuntu is hit or miss and has terrible software support (PPAs are annoying, and upgrades don't work well with them). I've been using Arch Linux with zero problems upgrading daily throughout the last two or three years. The only time you might have had an issue was when Arch switched to using systemd. If you are having issues with something breaking it's because you're not doing it right. Always perform pacman -Syu before installing new software, and it's always a good idea to log out and log back in after a lot of updates. I also use Gentoo which is also really great with package management.

i move from arch because in same week cinnamon desktop stop working(last year) twice because some gay push the new desktop without testing. I always refresh pacman and always reboot a machine after upgrades. I only talking about what happen with me nothing more. For some reason i don't use the ubuntu normal releases too

[sireangelus]

> A Few Worrisome Regressions Appear In Ubuntu 15.04 vs. 15.10 Performance

Yeah, right. Ubuntu is always in broken state. I guess we can call it "stable" as well....

Yeah right, because you are trying to use software that in their own admission is considered "beta" and unstable instead of just installing the stable one (14.04)

[geearf]

I'm a software engineer (degree, job title etc) and I've still experienced breakages when upgrading an Arch system. It's pretty lame that when they have a breaking update, there's no notice or other kind of warning in pacman. You really should check their website manually for such notices before every upgrade.

Actually, you subscribe to the ML, and you read the emails before using pacman and you're fine.
Though if you do daily ML like me, it might come a bit too late

[Dick Palmer]

Michael, would you be interested in comparing the problematic tests compiled with -O3 and -OFast in both releases to rule out some change to GCC's optimisation/flags?

PS Please stop feeding the obvious "troll" peeps

[linuxforall]

Ubuntu bashing starts while Windows watches and laughs and Apple joins in. Thats why there is LTS and please, I do run pure Arch and Gentoo in two separate machines but thats to teach and learn, not as my bread and butter machine. All of them including the 200 odd desktops and laptops as well as mainframe runs UBUNTU at my univ here setup by my students. Couldn't be better.

One can't run a running experiment in a work machine period.

[Michael]

Michael, would you be interested in comparing the problematic tests compiled with -O3 and -OFast in both releases to rule out some change to GCC's optimisation/flags?

PS Please stop feeding the obvious "troll" peeps

As with anything, if enough Phoronix Premium subscribers request such tests, this was all done out of the box. All the compiler flags was the same on each OS as shown in the results.

[SystemCrasher]

If you cannot upgrade Linux system then you should not be a Software engineer.

Software engineers would not waste their time on stuff like arch or gentoo. Engineer is one who achieves his goal, not one who just hits all knobs and buttons like mad "just because he can", like gentoo or arch scriptkiddies do. It's enough to take a look on "smartass" comments like we have got here to get very clear idea who is using arch and gentoo.

And btw, real software engineers would not tolerate system breakages in production. Yet its a hallmark of arch and gentoo. So, if we take a look on what actually being used in production, ubuntu beats arch and gentoo any time of day. You see, at least it got predictable timings of releases. So I know it would not break between releases and can pay extra attention once in a while on major upgrade which would require extra validation. It is good when system is predictable. OTOH gentoo and arch do not really care about locking software versions for a while, etc. Should author of program change something in incompatible way, something would break.

Isn't it cool to have latest version of your favorite httpd? Hmm, well, let's see. New version has arrived. With many new goodies. You've applied it. Whoops! It turns out authors decided that option you've used in config is now obsolete and removed. So httpd fails to start. Tee-hee. If you've been dumb enough to do it on important server, you have urgent emergency condition as bonus. Is it how engineers are supposed to keep production environments? :P

[geearf]

You do realize that you would not upgrade a production system without testing the upgrade on a non production system before?
(of course that's not a 100% check, but still you'd catch your example easily).
Same with predictable releases, you can upgrade at the pace you decide, so either way, it's predictable

(Now I wouldn't still use a standard Arch for production systems but your point is kind of weak).

[hwertz]

Could be the compiler. When I used to use gentoo... I've seen this before, I'd rebuild world with a newer compiler, they'd use a new optimizer or something and some stuff would run noticeably faster and some would slow down. Usually with a x.x.0 compiler version (say 4.6.0) then by x.x.2 or x.x.3 the regressions are fixed (it may not get any additional optimization compared to the old compiler but at least is no longer more poorly optimized. )