OS block size ?

[chris zeis]

I know that what the blocksize is for each of my volume groups. What is the OS block size?

I'm using RAID and know that I can determine the stripe sizes. The problem is when I request for example 256K worth of data, how does the OS handle the data and in what size?

[Kofi ARTHIABAH]

It depends on what (-b = block size ) parameter values you used when you created the filesystem. do a man on newfs_vxfs. The default for vxfs is 1024 bytes. (if you used SAM to create the fs the it is likely the default)

[Stefan Farrelly]

The block size of your filesystems is usually from 1k up. Use the fstyp command to see the blocksize.

To see the stripe size a simple lvdisplay will show what stripe size, if any, your lvols are using. If you are using RAID then this doesnt necessarily mean you have an lvol stripe size, your RAID disk device (eg. a Nike) may have its own internal stripe size when you bind disks together into a lun, you will need to look at the documentation for your RAID disks subsystem to deternine its stripe size, eg. a Nike can use from 8-128k.

All this block size info though doesnt really help you 'calculate' performance and what is optimal for your environment. Nothing beats doing your own tests. Do some raw dd's of your lvols and time them to see what the performance is like, modify the block size or stripe size and repeat the time dd to see whats best for your server/disks.
eg. time dd if=/dev/vgxx/rlvolxx of=/dev/null bs=1024k count=50

[James R. Ferguson]

Chris:

# fstyp -v /dev/dsk/cXtXdx

...JRF...

[John Palmer]

Have a look at this recent thread:-

http://my1.itrc.hp.com/cm/QuestionAnswer/1,1150,0xc9eb6c96588ad4118fef0090279cd0f9,00.html