-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi,

On Tue, 24 Jun 2003, David Griffiths wrote:

> I'm surprised there is not more interest in this; is it that not many
> work with large-ish (10+ gig) databases that need high-end performance?

Many of our customers do.

> A 64-bit CPU won't have the 4-gig memory limit that a 32-bit processor
> will; even worse, Linux is apparently limited to about a 2-gig process.

It depends - there is a BIGMEM patch from Andrea Arcangeli that raised
that limit to 3.5 GB on 32bit systems. I think the patch is in the
mainline kernel as well by now.

> SuSe Enterprise Linux supports 512-gigabyte processes with 16
> processors. Imagine 10 gigabyte database all in memory.
>
> Even better, larger file sizes - no more 2-gig files. Max file size is
> 9-Exabytes ( "9" followed by 18 "0's" ). All the posts I see about
> people trying to get around the 2 gig file limit should be really
> excitied.

Actually, you can create larger files on 32bit Linux systems as well. It's
just that the file system and the C library must have support for LFS
(Large File Support):

http://www.suse.de/~aj/linux_lfs.html

> I guess I'm just surprised by lack of interest. I've been bugging our
> CTO once a week about this, and hopefully should have a server on my
> desk by mid summer to late fall.

Have fun with it!

> If you're interested, SuSe has a good PDF on AMD64 and SuSe Enterprise Linux
> 1.0:
>
> http://www.suse.com/en/business/products/server/sles/misc/sles8_amd64.pdf
>
> Anyone have some practical experience with the software and hardware?

Yes, we do have two AMD64 systems (one dual Opteron with 1GB of RAM),
running SLES here. Quite impressive. We also have some Itanium (1/2)
systems and SGI kindly gave us access to an "Altix" system (16 x 900MHz
Intel Itanium-2, 32GB of RAM). So MySQL runs quite happily on 64bit
systems.

The main benefit of a 64 bit platform is that MySQL can handle more
concurrent threads and that you can give more than 4GB of memory to MySQL
buffers.

The disadvantage is that MySQL will be about 3-5 % slower because of the
extra memory usage 64 bit pointers require (Structures are aligned to be
64 bit aligned, which makes the structures bigger and reduces the CPU
cache usage, more stack space is used for registers). We tested this on
Sun Solaris, using the Sun Forte compilers, but I assume this applies to
other architectures as well.

Bye,
        LenZ
- -- 
For technical support contracts, visit https://order.mysql.com/?ref=mlgr
   __  ___     ___ ____  __
  /  |/  /_ __/ __/ __ \/ /      Mr. Lenz Grimmer <[EMAIL PROTECTED]>
 / /|_/ / // /\ \/ /_/ / /__     MySQL AB, Production Engineer
/_/  /_/\_, /___/\___\_\___/     Hamburg, Germany
       <___/   www.mysql.com
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.2-rc1-SuSE (GNU/Linux)
Comment: For info see http://quantumlab.net/pine_privacy_guard/

iD8DBQE++VFtSVDhKrJykfIRAsLHAJ4nx0SfyJxBhtMQW+nodnjXNArSdQCeKxL8
G6ScGyAcT8tN4mI++T1K7j4=
=7ti/
-----END PGP SIGNATURE-----

-- 
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:    http://lists.mysql.com/[EMAIL PROTECTED]

Reply via email to