[CentOS] athlon64/opteron 8GB per CPU

Mon Apr 10 19:51:13 UTC 2006
Jay Lee <jlee at pbu.edu>

David Thompson wrote:
> I'm putting together a spec for a big(ger) memory x86_64 host, and I hope some 
> of you-all can help.  I'm looking for success stories of hardware that folks 
> are using in production environments, preferably with CentOS 4, with more than 
> 4GB of memory per CPU.
I have a dual opteron server with 6gb ram running RHEL4u3 (not CentOS 
but you understand).  It has an Adaptec SCSI RAID controller and 5 70gb 
drives in a RAID5 array (mailserver).
> I know there are lots of mobos out there that can do this, but I'm looking for 
> what folks really have running.  Call me paranoid.
Paranoid nothing, I fully understand the desire to get recommendations 
on a board, I've been burned in the past also.  The board we're using is 
a Tyan S2882.  They are not dual core Opteron's however we have a 
similar server running a newer Tyan board, 16gb ram and dual core 
opterons.  It's been quite stable on the Operating System it runs (I'll 
give you a hint, it comes from Redmond, WA)
> The target is SATA + n*GigE + 1CPU + 8GB memory, and I can consider other 
> configurations.  The incumbent system is a Xeon 8GB system running i686 CentOS 
> 4.3 (what else?), with light CPU loading and very heavy IO loading (WD raptors 
> are our friends).
Why SATA?  Are you using the onboard controller?  You may have issues 
here.  I'd at least consider a SCSI RAID card if it's in the budget.  If 
not, be sure your board's SATA controller is well supported.

Why i686?  Sure the i386 version of CentOS 4 can access >4gb ram but 
it's quite ugly how it has to do it and there are still some limits.  
I'd strongly recommend using the x86_64 release if you can pull it off.

> Folks with deep(er) knowledge of x86_64 architecture, I'd also be interested 
> in the trade offs between 1CPU w/8GB and 2CPUs each with 4GB.  (E.g. what's 
> the real cost of shipping data back and forth between the memory spaces of the 
> two CPUs?  How does the 2.6 kernel deal with the split memory space?  etc...)  
> The memory is all going to be used up in kernel buffers keeping bits of files 
> in kernel memory.
Here's where the Opteron with onboard memory controllers will kick 
Xeon's rear.  Also, throwing more CPUs in the mix is not going to cut 
down on the amount of memory available, a system with 2-cpus and 8gb 
total ram will always be faster than a system with 1 cpu and 8gb ram 
(except in really strange workloads).
> Also, trade offs between Opteron and Xeon architectures, i686 vs. x86_64, 
> bounce buffers, etc., for an application like this would be helpful.
Opteron rules.  Their x86_64 support is better and the onboard memory 
controller is much faster.  Also, the Opterons have dedicated pathways 
between CPUs which means much faster process sharing and communication 
between CPUs.  As I understand it if CPU1 has an area of memory cached 
and CPU2 needs that data, it can grab it from CPU1 directly.  Xeons 
would need to go back to RAM to grab the data which is much slower.
> The application is enhancement to a mirror server (mirror.cs.wisc.edu) here at 
> UW-Madison.
Well, if all you're doing is rsync or similar than CPU really isn't 
going to be a bottleneck in the near term, bandwidth and I/O are the 
bigger bottlenecks.  What speed will the network be running at?
> Many thanks to all in advance.  Private or public replies welcome.
> Dave Thompson
> UW-Madison
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> http://lists.centos.org/mailman/listinfo/centos

-------------- next part --------------
A non-text attachment was scrubbed...
Name: jlee.vcf
Type: text/x-vcard
Size: 193 bytes
Desc: not available
URL: <http://lists.centos.org/pipermail/centos/attachments/20060410/236daff7/attachment-0005.vcf>