A 10-Way Linux Distribution Battle To Kick Off 2016

[darkbasic]

Fixed that for you. Even the Debian Testing Security team admits that security support for Testing is not as good as for Stable.

Don't waste your time answering, he clearly never administered a real production server.

[blackiwid]

very confusing tests way bigger differences than I thought, but many tests are very situational for some devs compile times are everything for most people its nearly irrelevant, same is true for apache benchmarks... for desktop user nearly irrelevant...

So a middled speed rating would help not at all, you have to know what is most important for you and how fast is the distro at this task.

its also not very desktop centric, there a sunspider benchmark or something would be more interesting.

hmm k now I get the biggest problem of this benchmark:

"Fedora 23 currently has the Linux 4.2 kernel,"

I am ot shure this statement is true or did the benchmark just missed by a day or a few the update:

kernel x86_64 4.3.3-300.fc23 updates 50 k
kernel-core x86_64 4.3.3-300.fc23 updates 20 M

Repo : updates

is update not official? k maybe I get there the kernel update a weak earlier, shurly not much more.

[edmon]

Don't waste your time answering, he clearly never administered a real production server.

point me one security problem with debian testing in last 15 years(in which i am using it in production) which can lead to whatever harm to the server and i will believe you or you have to delete your previous messages.

[Michael]

"Fedora 23 currently has the Linux 4.2 kernel,"

I am ot shure this statement is true or did the benchmark just missed by a day or a few the update:

kernel x86_64 4.3.3-300.fc23 updates 50 k
kernel-core x86_64 4.3.3-300.fc23 updates 20 M

Repo : updates

is update not official? k maybe I get there the kernel update a weak earlier, shurly not much more.

Considering it takes several days to run all the benchmarks, may have missed the new kernel on Fedora by a day or so.

[darkbasic]

point me one security problem with debian testing in last 15 years(in which i am using it in production) which can lead to whatever harm to the server and i will believe you or you have to delete your previous messages.

I never mentioned security problems myself, the main issue with testing is the maintenance burden compared to stable. The whole rolling release concept is completely wrong for servers.

[carewolf]

Looking at the Intel Linux compiler flags, it looks like it boils down to SSE4.2 and -O3. The O3 enables autovectorization which can double or quadruple performance when hitting things it works own, and with SSE4 the things that can be autovectorized is improved.

So the packages that benifit from it should just enable it as default.

[jmcknight]

No, for desktops, testing has plenty of uses, especially if you don't want to be stuck with ESR browsers (which update 7 times less often than regular ones).

If you don't need high availability, using testing on a server is perfectly acceptable.

Or people who require serious stability and don't mind using slightly older versions of things. There are people who don't need or want the latest of everything because the latest doesn't always equate to stablest. I use it for the same reason that people use CentOS and SLED except without RPM awfulness.

[Kano]

@Calinou

You don't need to use ESR versions for Debian, you can simply use:

http://mozilla.debian.net/

I use that for Iceweasel (Firefox rebrand) for Kanotix, but not for Icedove (Thunderbird) because i don't think it matters that much there.

[lowflyer]

An interesting follow-up test would be to compare these results with the same results from an AMD system of comparable performance. The ClearLinux distribution is aimed "to showcalse the best of Intel Architecture technology". How well does ClearLinux perform on AMD compared with other more "processor agnostic" distributions.

[sanne]

I understand that Michael can't spend too much time on this, and tuning the machines "correctly" would take a huge amount of time (and expert help needed), but it would be great to compare these systems as a skilled admin would tune them. I don't think that - if you care about performance - you'll run with out of the box settings?

Especially the Red Hat based distributions, I'm sure there's awesome tuning options for the database benchmarks just by configuring the DBs and enabling the right tuned profile.

Also I would love to see stuff like Arch and Gentoo included in such comparisons. I always wondered if it's worth the time to compile the whole system with options like "march=native".. I might give Gentoo another shot myself, but I can't possibly wipe my system several times to do a full comparison with all these benchmarks.. I have to use my workstation for actual coding ;-)