NAT Server, SMP box or Single CPU box?

NAT Server, SMP box or Single CPU box?

[Donato Diaz]

[-- Attachment #1: Type: text/plain, Size: 453 bytes --]

Hi,

I have read in different lists that some people are having problem doing 
NAT with SMP Servers, due to the way the kernel manage conntrack with 
both CPUs.
NAT runs slower on Server SMP with two CPUs?
Is it better a Server with one faster CPU than a server with two CPUs?
I need to buy a NAT Server for +1500 users. I would like to know if an 
SMP Server with two CPUs is the best choice.

Any help would be really appreciated.
Thanks all,
Donato.

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3018 bytes --]

NAT Server Performance, SMP box or Single CPU box?

[Donato Diaz]

Hello,
I have read in different lists that some people are having problem doing 
NAT with SMP Servers, due to the way the kernel manage conntrack with 
both CPUs.
NAT runs slower on Server SMP with two CPUs?
Is it better a Server with one faster CPU than a server with two CPUs?
I need to buy a NAT Server for +1500 users. I would like to know if an 
SMP Server with two CPUs is the best choice.

Any help would be really appreciated.
Thanks all,
Donato.

Re: NAT Server Performance, SMP box or Single CPU box?

[Harald Welte]

[-- Attachment #1: Type: text/plain, Size: 954 bytes --]

On Wed, Dec 29, 2004 at 10:15:21AM -0300, Donato Diaz wrote:
> Hello,
> I have read in different lists that some people are having problem doing 
> NAT with SMP Servers, due to the way the kernel manage conntrack with 
> both CPUs.

this isn't a development question. Please follow-up to
netfilter@lists.netfilter.org

> NAT runs slower on Server SMP with two CPUs?

no, but the gain is not something like 'twice the performance'.  

with networking, your i/o subsystem (64bit, pci-x, fast RAM, ...) is way
more important than cpu speed.

> Thanks all,
> Donato.

-- 
- Harald Welte <laforge@netfilter.org>             http://www.netfilter.org/
============================================================================
  "Fragmentation is like classful addressing -- an interesting early
   architectural error that shows how much experimentation was going
   on while IP was being designed."                    -- Paul Vixie

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]