[Beowulf] Gentoo in the HPC environment

Gerald Henriksen ghenriks at gmail.com
Sat Jun 28 12:53:04 PDT 2014


On Sat, 28 Jun 2014 15:38:22 +0200, you wrote:

>My arguments for gentoo is more along the lines of optimization of the
>source code for ones hardware.

Every now and then someone gets on the Fedora list
suggesting/demanding that they change some compiler flags or other
optimizations that the person has used on a distribution like Gentoo
and seen "significant" performance gains.

The standard response is for the person to do some benchmarks that
prove the claimed improvement is real, and then it will be considered.

So far, no one has come back with proof of improvements.

For most use cases the OS is not a limitation, and expending effort to
improve things will not gain any significant improvements and may
possibly make things worse.

>You mention CentOS and Red hat how optimized is the heart of the system
>for the hard ware its running on?

What is there to optimize?

Yes, the kernel and a few key libraries have internal code paths that
can optimize based on processor but the 64bit world is a fairly
uniform instruction set (unlike the 32bit Intel world) with the
exception being the SSE and AVX type extensions which aren't of much
use to the OS.

>I for instance currently even though nothing is running on it have a vps
>with 2gb of ram and with no websites or db on there yet just the servers
>ram wise i use about 62mb. I look at the low memory foot print as an
>advantage in the sense that the rest of the ram can be used for more
>important things.

That isn't so much optimizing for hardware as it is making sure you
are only running services/daemons/programs that you need for your
requirements, which is easy to do with any distribution.


More information about the Beowulf mailing list