SCIENTIFIC-LINUX-USERS Archives

November 2005

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:
Troy Dawson <[log in to unmask]>
Reply To:
Troy Dawson <[log in to unmask]>
Date:
Tue, 8 Nov 2005 08:56:20 -0600
Content-Type:
text/plain
Parts/Attachments:
text/plain (68 lines)
Miles O'Neal wrote:
> Any idea how to tell if an SL3 kernel
> has NUMA support enabled?
> 
> Thanks,
> Miles

You could read the release notes.  These are found at
/usr/share/doc/sl-release-3.0.5/
or
ftp://ftp.scientificlinux.org/linux/scientific/30x/x86_64/SL.documentation/
ftp://ftp.scientificlinux.org/linux/scientific/30x/i386/SL.documentation/

I would suggest using the local copy, because it's easier to grep 
through them.  But if you don't have the latest release, then the online 
copy will do.

grepping through them, then only place with NUMA is in the Update 5 
x86_64 release notes, which say

--------------------------
    This section contains information related to the Scientific Linux
    3.0.5 kernel.

   o Systems with processors featuring AMD64 and Intel(R) EM64T are
   typically configured as NUMA platforms, which means that the kernel
   constructs multiple memory nodes at boot-time rather than constructing
   a single memory node. The multiple node construct can result in memory
   exhaustion on one or more of the nodes before other nodes become
   exhausted. When memory exhaustion happens, the following could result:

   1) The system will swap the exhausted nodes while there is available
   memory on other nodes, resulting in poor overal performance

   2) Processes are killed due to Out-Of-Memory (OOM) errors even though
   there is available memory

   3) Less than optimal performance due to excessive memory bandwidth
   when processes running on an exhausted node allocate memory on one or
   more different nodes

   The kernel is capable of detecting when OOM process kills occur on
   AMD64 and Intel(R) EM64T NUMA systems due to memory exhaustion on one
   or more nodes while there is available memory on other nodes. When the
   kernel detects this condition, it displays the following warning at
   the console:


OOM kill occurred on an x86_64 numa system! The
numa=off boot option might help avoid this.

   Adding numa=off at the boot command line in /boot/grub/grub.conf and
   rebooting the system disables the creation of separate memory nodes at
   boot-time and elminates the possibility of one memory node becoming
   exhausted even though other nodes have available memory.

-------------------------------

I know this doesn't tell you generically how to check, but it's good 
stuff to know if you're wondering about NUMA in S.L. 3.x

Troy
-- 
__________________________________________________
Troy Dawson  [log in to unmask]  (630)840-6468
Fermilab  ComputingDivision/CSS  CSI Group
__________________________________________________

ATOM RSS1 RSS2