SCIENTIFIC-LINUX-USERS Archives

July 2011

SCIENTIFIC-LINUX-USERS@LISTSERV.FNAL.GOV

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
夜神 岩男 <[log in to unmask]>
Reply To:
夜神 岩男 <[log in to unmask]>
Date:
Mon, 25 Jul 2011 17:39:40 +0900
Content-Type:
text/plain
Parts/Attachments:
text/plain (42 lines)
On 07/25/2011 04:45 PM, Artem Trunov wrote:
>         Hi
>
>
>     I am not aware of any actual test data that compares the various
>     RHEL derived distros under any stress in a meaningful way (are you
>     volunteering?).
>
>
> My colleges benchmarked SL60 vs. Ubuntu 11.04 server using Phoronix and
> in most of the tests U went ahead of SL by few percent. I got curious
> about this and tried to tune something - recompiled stock kernel with
> some server performance options like config_preempt_none, and also used
> recent fedora kernel 2.6.38 to match that of U. While this brought some
> marginal improvements in SL test scores, U was still ahead.
>
> Don't know how to explain this apart from that RH6 is older that U11.04
> or that ubuntu kernel is server optimized (build options at least), but
> still interested to hear some ideas from experts...

Benchmarking for speed is not what I think the OP had in mind.

But on the note of speed benchmarking, vanilla Ubuntu usually is a bit 
faster than vanilla Fedora/RHEL because Ubuntu does not enable SELinux 
by default. The benefits of SELinux are considerable, but it results in 
a performance penalty which varies with the type of tasks you are 
putting the system to in the test. Tests with lots of filesystem access 
requests tend to hit a several % point penalty because of the extended 
reqirements SELinux entails, whereas this effect is minimalized when 
comparing similar data-fetching but within, say, an RDBMS like PostgreSQL.

Generally the speed differences are marginal and offset against security 
gains, but environment and application really dictates this. Anyway, 
because of these kinds of default differences between Debian and Fedora 
based systems figuring out what is causing speed to vary where is a 
little tricky and can go beyond simple examination of the kernel build. 
There are a *lot* of variables -- and in practical applied scenarios 
they go way beyond SELinux (dash vs bash, filesystem options, etc. there 
are lots of tiny differences in approach on both).

-Iwao

ATOM RSS1 RSS2