how to know the block size and chunk size of my hp ux system

[Eric Unix]

Hello All

Three question need your help
1. how to know the block size of my hp ux system ?

2. what is the meaning of " chunk size " ?

3. how to know the size of chunk size of my hp ux system ?

Appreciate all of your good answers

BR
eric

[candlejack]

1. I assume you're wanting to know because you want your Oracle blocksize to match. Run:

fstyp -v /dev/rdsk/

Sorry, don't know about chunksizes.

[Hein van den Heuvel]

Eric,

Did you try a quick Google on +hpux +blocksize
Optionally add +site:itrc.hp.com

Many folks have asked this before, most for the wrong reasons. Many answer have been provided.

>> 1) how to know the block size of my hp ux system ?

Hp ux systems do not have a block size (or many). Filesystems and storage units do.
Probably you refer to a FILESYSTEM?
If so, WHAT filesystem? JFS? HFS

As indicated the most likely answer will come from : /usr/sbin/fstyp -v

8192 bytes... mostly. 1024 sometimes.
4096 is a nice number round number too!

512 byte may be the smallest block you can read or write.

>> 2) what is the meaning of " chunk size " ?

In what context? cat-food?

Typcically it refers to the a smart IO controller using 'chunks' of blocks to implent its RAID algoritmes. This if often implented TRANSPARENT to the OS.
Ideally it is chosen based on anticipated load patterns. Often, a default is accepted or the storage admins favorite number is used and it migth take 'an act of god' to change it.

It might refer to the LVMS PHYSICAL EXTEND size.
Check out 'man vgcreate', notably the -s option.

>> 3) how to know the size of chunk size of my hp ux system ?

You nicely ask the person responsible for setting up the storage.
As I indicated, the chucnk size is something the raid controller deals with and may keep entirely transparent to the storage consumers. To to know the chucksize you may have to ask the storage controller. How to ask that, is controller dependent. XP? EVA? DMX?


What problem are you really trying to solve?

What do you know about your (storage) environment?
- LVM?
- RAW or Filesystem?
- Direct-acttached disk or smart/raid controller? Which one?

Good luck,
Hein van den Heuvel
HvdH Performance Consulting.

[Prashanth Waugh · Accepted solution]

Hi Eric

#df -g /tmp

will show the block size

Regards
Prashant

[Prashanth Waugh]

Hi Eric,

As per my knowledge chunk is nothing but the block size. while creating the FS we r mentioning the block size. u know that tape and disk are block devices means while taking the backup in tape it will take block by block .So we can conclude that block size is chunk size.

Regards
Prashant

[Deepak Kr]

Hi,

You can find block size in hp-ux using :

#fstyp -v /dev/vg00/lvol6
vxfs
version: 4
f_bsize: 8192 ====>block size
f_frsize: 8192
f_blocks: 65536
f_bfree: 60292
f_bavail: 60077
f_files: 10752
f_ffree: 15072
f_favail: 15072
f_fsid: 1073741830
f_basetype: vxfs
f_namemax: 254
f_magic: a501fcf5
f_featurebits: 0
f_flag: 16
f_fsindex: 7
f_size: 65536

Also:

vgdisplay vgname |grep "PE Size"

As Hein said chunksize is more relevent to Storage area..

In RAID under storage Stripes go across disk drives. But how big are the pieces of the stripe on each disk? The pieces a stripe is broken into are called chunks


chunksize is very important thing when we are designing/using RAID architecture.

[Eric Unix]

Dear All

Thanks in advance for all of good answers. Yes, I got more clear of it.
Yes, I want to map them with ORACLE's some parameters....
Log_buffer and sort area size etc ..

Best Regards
Eric

[Hein van den Heuvel]

Well, in that case...

The blocksize should not matter at all, as Oracle should be configured to bypass file system IO either through DIRECT-IO or RAW devices.

And the chunksize whilest very intersting is of limited usage. Check out your Oracle stats... are you not mostly doing single db-page IO? (sequential) if so... those will fit in any chunk.
It _might_ have a small dependency on multi-block read count for tablescans and such, but there you do not control the start page, so being too smart could backfire (always mis-align).

The log buffers IO is strictly variable length... whatever has been accumulated is pushed out, and you just have to make the buffer large enough.

In all, IMnsHO there are much bigger fish to fry, where you have more control and more chance for positive results.

With no disrespect intended, the nature of the question and data provided, suggest you are not 'ready', by a long shot to try to tweak this.

Absolutely check out the chunk-size and talk to the storage folks, but don't expect to be able to do much about it.

Best regards,
Hein van den Heuvel
HvdH Performance Consulting