Block Size

[sibsbt]

Setting Block Size in VMFS partitions

How can i change the  VMFS Block Size ?

The Block Size is set 8 MB on VMFS Data Store

At the Same time  on Storage is set to 128 KB( IBM SAN)

And The os Level (windows NTFS ) it is 4096 Bytes

In the Above scenario What will happen when a  guest VM doing a write operation

Please Explain

[rickardnobel]

sibsbt wrote:

How can i change the  VMFS Block Size ?

It can not be changed. You would have to migrate all vmdk files away, delete the datastore and then re-create to select something else.

sibsbt wrote:

The Block Size is set 8 MB on VMFS Data Store

At the Same time  on Storage is set to 128 KB( IBM SAN)

And The os Level (windows NTFS ) it is 4096 Bytes

In the Above scenario What will happen when a  guest VM doing a write operation

If the application in the guest issues a 512 byte write then a 512 byte write IO request will travel from the operating system to the disk. Since the disk is now a virtual disk the VMkernel will send this IO request to the actual disk area on the SAN, still a 512 byte write request. What will happen on the SAN is dependent on the vendor, so you will have to check this specific.

[sibsbt]

As Far as i know  when  the OS write 512 bytes the  Storage (Segment Size 128 KB) it equally divide the data  (4 Chunks ) and  write in to the sector .

And the same time  if the  the data is less than the segment size (say 1 KB ) , it still occupy 128 KB instead of 1 KB.

I just want to know   when  guest vm write 1 KB how  will be the flow .

Second thing if  we increase or decrease the block size of vmfs datastore , what will be the impact

[rickardnobel]

sibsbt wrote:

I just want to know   when  guest vm write 1 KB how  will be the flow .

From the Vmkernel side this will generate a 1 KB write request to the SAN, which then will be handled different depending on the SAN vendor. There is no relation with the IO size and the VMFS block size from the ESXi point of view.

sibsbt wrote:

Second thing if  we increase or decrease the block size of vmfs datastore , what will be the impact

The biggest impact is the maximum file sizes for VMDK, which depends on the block size.

[TomHowarth · Accepted solution]

sibsbt wrote:

Setting Block Size in VMFS partitions

How can i change the  VMFS Block Size ?

The cannot be changed once the datastore has been formatted without a re-format of the datastore

The Block Size is set 8 MB on VMFS Data Store

At the Same time  on Storage is set to 128 KB( IBM SAN)

And The os Level (windows NTFS ) it is 4096 Bytes

In the Above scenario What will happen when a  guest VM doing a write operation

Please Explain

it is best to refer to your SAN Vendor for this.

[TomHowarth]

sibsbt wrote:

Second thing if  we increase or decrease the block size of vmfs datastore , what will be the impact

you will not be able to create as big VMDK's for example if you move down from 8MB block size to a 1MB block size your maximum VMDK size will be 256GB rather than 2048GB - 512KB

[trink408]

I've often wondered about this block size question.

So what is the advantage of creating smaller block size, say 1mb block size vmfs datastores? If their is no advantage to it, then why not just always create 8mb block size so you have the greatest level of expansion?

Thanks

Kevin

[a_p_]

So what is the advantage of creating smaller block size, ...

There might be no advantage. However, keep in mind that e.g. migrations will take longer if the source and target datastore differ in block size. Also think of ESXi 5, where block size is no issue anymore since it uses a unified block size of 1 MB for all file sizes.

See some interesting discussions at e.g.:

http://www.yellow-bricks.com/2011/02/18/blocksize-impact/

http://www.yellow-bricks.com/2009/05/14/block-sizes-and-growing-your-vmfs/

(don't miss to read the comments too)

André