Very big shm area

Very big shm area

[Kilobug]

Hello,
	I wanted to know if it is possible to have a very big system V shared 
memory segment (say about 1Gb) ?

	I've quickly looked into the source code of shm.c and shm.h in ipc/ and 
I've read the following:
/*
  * SHMMAX, SHMMNI and SHMALL are upper limits are defaults which can
  * be increased by sysctl
  */

But how far is it possible to increase them ? And which sysctl must be 
done ?

Thank you for answering,

-- 
** Gael Le Mignot "Kilobug", Ing3 EPITA - http://kilobug.free.fr **
Home Mail   : kilobug@freesurf.fr          Work Mail : le-mig_g@epita.fr
GSM         : 06.71.47.18.22 (in France)   ICQ UIN   : 7299959
Fingerprint : 1F2C 9804 7505 79DF 95E6 7323 B66B F67B 7103 C5DA

"Software is like sex it's better when it's free.", Linus Torvalds

Re: Very big shm area

[Austin Gonyou]

I run a Dell 6450 with 8GB ram and a 4GB SHMMAX. You can set the
parameter any time through sysctl if you have it enabled, or you can
hardcode it per the Oracle/DB2 Docs if you wish. 

Austin


On Sun, 2002-06-02 at 12:21, Kilobug wrote:
> Hello,
> 	I wanted to know if it is possible to have a very big system V shared 
> memory segment (say about 1Gb) ?
> 
> 	I've quickly looked into the source code of shm.c and shm.h in ipc/ and 
> I've read the following:
> /*
>   * SHMMAX, SHMMNI and SHMALL are upper limits are defaults which can
>   * be increased by sysctl
>   */
> 
> But how far is it possible to increase them ? And which sysctl must be 
> done ?
> 
> Thank you for answering,

Re: Very big shm area

[William Lee Irwin III]

On Sun, Jun 02, 2002 at 07:21:21PM +0200, Kilobug wrote:
> Hello,
> 	I wanted to know if it is possible to have a very big system V 
> 	shared memory segment (say about 1Gb) ?
> 	I've quickly looked into the source code of shm.c and shm.h in ipc/ 
> 	and I've read the following:
> /*
>  * SHMMAX, SHMMNI and SHMALL are upper limits are defaults which can
>  * be increased by sysctl
>  */
> But how far is it possible to increase them ? And which sysctl must be 
> done ?
> Thank you for answering,

ls /proc/sys/kernel/sh* and the names shouldn't be too tough from there.
I'd be concerned about exercising this code on a virgin 2.4.17 as ISTR
bugfixes for sysv ipc/shm cropping up in later kernel versions, though
I can't say I've tracked this area very closely. Hopefully someone who
does can chime in here and enlighten us both.


Cheers,
Bill

Re: Very big shm area

[Austin Gonyou]

To get *good* performance from such a large SGA, it's best to use a -aa
kernel I've found. Either 2.4.18-aa + whatever patches, or something
similar. But there's been some fixes/changes/optimizations to the -aa
tree with regards to large memory boxes. 

I'm curious though how P4 HT will play into that performance on the new
Xeons though. I'll have more to report in the next few weeks.


On Sun, 2002-06-02 at 13:37, William Lee Irwin III wrote:
> On Sun, Jun 02, 2002 at 07:21:21PM +0200, Kilobug wrote:
> > Hello,
> > 	I wanted to know if it is possible to have a very big system V 
> > 	shared memory segment (say about 1Gb) ?
> > 	I've quickly looked into the source code of shm.c and shm.h in ipc/ 
> > 	and I've read the following:
> > /*
> >  * SHMMAX, SHMMNI and SHMALL are upper limits are defaults which can
> >  * be increased by sysctl
> >  */
> > But how far is it possible to increase them ? And which sysctl must be 
> > done ?
> > Thank you for answering,
> 
> ls /proc/sys/kernel/sh* and the names shouldn't be too tough from there.
> I'd be concerned about exercising this code on a virgin 2.4.17 as ISTR
> bugfixes for sysv ipc/shm cropping up in later kernel versions, though
> I can't say I've tracked this area very closely. Hopefully someone who
> does can chime in here and enlighten us both.
> 
> 
> Cheers,
> Bill
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/