LVM disk overhead

[Joseph Hoh]

What % of the disk does LVM config info. use. Can I reduce this so I can use more of the physical disk space?

[Andreas Voss]

Hi,

available disk space is not dramatically reduced by LVM.
The major thing is the filesystem on it.
For hfs you can reduce the minfree value at fs creation time with: newfs -F hfs -m 1 ...
(the default is 10%, example above uses 1%)
For vxfs this is nt available.

Regards

[federico_3]

For hfs file systems the % used by LVM is about 10% ot the allocated space (note the difference between the allocated space from bdf FS
df -k FS).
For vxfs ones the % is lesser.

I think you cant reduce this !

Federico

[Stefan Farrelly]

You can save quite a lot of space on VXFS filesystems, especially if they are large (as more space is reserved within them which may be unnecessary if you only have a few large db files on it). You need to use the vxtunefs command, part of;

B3929CA B.03.03 JFS 3.3 Filesystem For 11.00

[Carlos Fernandez Riera]

You can make phisiscal extension largest, for example 8Mb, and so use fewer space when pvcreate, but this would affect performance.

I have see good explanation about PE size, disk space and performace in this forum , but i cant find it now. Try search.

Changing minfree would affect performance too, but, for example, i change usualy this parameter is some filesystems.


As always, change defaults depends of your needs ( usualy urgent needs). Analyze posibilitys and needs before do it.

[Joseph Hoh]

Where do I find the JFS 3.3 Filesystem fo 11.00?

How do you determine how much you can reduce the overhead?

How much overhead does LVM take on a vxfs filesystem?

[Stefan Farrelly]

You have to purchase JFS 3.3 from HP.

For more info on LVM overheads check out the manual which comes with JFS 3.3 or docs.hp.com

[John Palmer]

Hi Jeff,

LVM only takes a few Megabytes to hold its information about the disk.

It doesn't impose any overhead on a filesystem, only on the disk itself - to hold details of the volume group itself such as what logical volumes are where etc.

The biggest overhead on a filesystem (HFS or VXFS) is the number of i-nodes that are created. The default for a VXFS filesystem is about 6% (from memory). Try creating a filesystem with the default number of i-nodes then recreate it with a specified number (mkfs -o ninode=n...) and you'll see what I mean.

Have a look at this thread:-

http://my1.itrc.hp.com/cm/QuestionAnswer/1,1150,0xec707e990647d4118fee0090279cd0f9,00.html

Regards,
John

[Alan Riggs]

John hit it on the head (surprise!).

The biggest overhead for large filesystems is in inode allocation. The default inode allocation is almost always much larger than necessary, especially for database filesystems.