numa question for xeon 5500

On Feb 13, 3:12*pm, Yousuf Khan wrote:
Sven Bächle wrote:
Thanks for all the answers. No, it's not a performance issue, at least
not at the moment. I was just unsure wether it was possible, but it
makes sense that the MMU and OS take care of that.
Raising the question of performance - is this true:

* * *More importantly, I should point out that even
* * *remote memory references in Nehalem-based Servers
* * *are faster than all memory references in the
* * *previous generation Xeon-based systems.

I read it on
http://kevinclosson.wordpress.com/20...numa-system-or...

Another thing to note, is that in this article that you pointed to, they
are talking specifically about Oracle database performance. When NUMA
first started becoming a big issue back in 2003, when the Opteron first
came out, AMD coined the term SUMA (Sufficiently Uniform Memory Access).

Your retelling of history as you invariably misremember it. For one
thing, you are conflating NUMA with serial interconnects and
discarding the front-side bus. SGI used cache coherent NUMA-Link with
Itaniums that most certainly used a front-side bus. Even in the case
of cache coherent serial interconnects, others were there firstest
with the mostest.

http://it.toolbox.com/wiki/index.php/NUMA_Architecture

That Intel got away with a FSB for so long is the marvel, unless you
want to give credit to AMD for (as usual) claiming credit for copying
what others have already developed.

Robert,

* Yousuf Khan:
When NUMA
first started becoming a big issue back in 2003, when the Opteron first
came out, AMD coined the term SUMA (Sufficiently Uniform Memory Access).
That means just treat the memory as a big flat-space ignore the NUMA. At
that point in time, the operating systems available were Windows 2000/XP
and Server 2000/2003, mainly 32-bit versions of them too. They weren't
too NUMA aware, though they had some rudimentary awareness.

That's not true. Windows Server 2003 (Enterprise/Datacenter Edition) is
fully NUMA-aware, both in the 32bit and in the 64bit version. Windowsxp
Professional 32bit does support NUMA at least since SP2, and since
Windowsxp x64 basically is Windows Server 2003 it does support NUMA,
too. Vista, Windows Server 2008 and Windows 7 of course also have full
NUMA support. All these Windows versions have static NUMA support, there
however are (especially in the desktop and smaller server versions of
Windows) certain limits in support of dynamic ressource changes (i.e.
changing the amount of RAM and no of CPUs at runtime).

However, NUMA support relies on the provision of a valid SRAT (Static
Ressource Allocation Table) by the underlying BIOS. Especially many
older Socket 940 Opteron machines didn't provide a SRAT.

Benjamin

On Feb 18, 1:09*am, Benjamin Gawert wrote:
* Yousuf Khan:

When NUMA
first started becoming a big issue back in 2003, when the Opteron first
came out, AMD coined the term SUMA (Sufficiently Uniform Memory Access)..
That means just treat the memory as a big flat-space ignore the NUMA. At
that point in time, the operating systems available were Windows 2000/XP
and Server 2000/2003, mainly 32-bit versions of them too. They weren't
too NUMA aware, though they had some rudimentary awareness.

That's not true. Windows Server 2003 (Enterprise/Datacenter Edition) is
fully NUMA-aware, both in the 32bit and in the 64bit version. Windowsxp
Professional 32bit does support NUMA at least since SP2, and since
Windowsxp x64 basically is Windows Server 2003 it does support NUMA,
too. Vista, Windows Server 2008 and Windows 7 of course also have full
NUMA support. All these Windows versions have static NUMA support, there
* however are (especially in the desktop and smaller server versions of
Windows) certain limits in support of dynamic ressource changes (i.e.
changing the amount of RAM and no of CPUs at runtime).

However, NUMA support relies on the provision of a valid SRAT (Static
Ressource Allocation Table) by the underlying BIOS. Especially many
older Socket 940 Opteron machines didn't provide a SRAT.


Yes, indeed. Weenie details rule.

Robert.