Thin and Wide
by Hu Yoshida on May 27, 2007
Most of the posts are about the benefit of reducing allocated but unused capacity.
There are some other benefits that should not be over looked.
The HDS approach to thin provisioning is to create a storage pool (Hitachi Dynamic Provisioning or HDP pool)which contains many logical devices. Capacity is allocated to “thin” volumes from this pool in units of 42 MB pages which are sliced across all the logical devices in the pool.
A storage pool can consist of tens to hundreds of logical devices. A page then consists of a wide stripe that spans tens to hundreds of logical devices which can be used to support an I/O request. This can have a dramatic increase in performance. It will also reduce the administrative effort that is required to tune volumes that are allocated to a limited number of logical devices since the volume will be automatically striped across all the logical devices in the storage pool.
Some users may even opt to assign one volume to one storage pool, just to get the performance benefit of wide striping, and not worry about the allocated but unused space.
The USP can support 32 storage pools today with plans to increase it in the future.
Hitachi Dynamic Provisioning can be viewed as two products, a thin provisioning product as well as a wide striping product for automated performance tuning..
Comments (2 )
This blog is hard to find on our external hds.com site. I had to do a search under blog.
Could you ask the web marketing team to highlight your blog on the landing page?
I think I understand how Thin Provisionig works, but I can´t find any information about how zfs and thin provisioning work togheter. I just want to know which percentage of the “promised” LUN is actually allocated when we assigned this luns to a zfs poll and define a zfs filesystem in it.
Any answer of you will help me a lot.
Thank you in advance.