Provisioning Thin Volumes

The capacity that is reserved for thin provisioning, which is a part of the system overhead, is accounted for in the available capacity that Oracle FS System Manager (GUI) reports. Because the system has already accounted for this reserve capacity, what the GUI reports as available capacity is fully available for the provisioning of logical volumes.

For SAN LUNs, the degree to which a LUN is thinly provisioned depends on the nature of the host applications that access the LUN. If only specific portions of a LUN are ever accessed by applications, the thinness of that volume remains the same. As applications attempt to access increasingly diverse areas of the LUN, however, the Oracle FS System allocates additional physical space for the LUN, causing the thinness of the volume to decrease.

Some applications access most or all of the addressable space for a volume. In these cases, the volume transitions from being thinly provisioned to being fully provisioned while the application executes.

The Microsoft Windows operating system reserves a substantial amount of metadata for a filesystem that has been formatted as a New Technology File System (NTFS) volume. The layout of this metadata causes an early allocation of thinly provisioned space. The primary NTFS metadata consists of the following objects:

To prevent the MFT from becoming fragmented, Windows reserves a buffer around the MFT. The size of this buffer can be configured to be 12.5%, 25%, 37.5%, or 50% of the drive space. Windows does not create new files in this buffer region until the unused space is consumed. Each time the remainder of the drive space becomes full, the buffer size is halved. This strategy provides new space for additional write operations.

Oracle does not recommend creating a thinly provisioned LUN that consumes more than 90% of the maximum capacity on the first infill, especially when using NTFS. NTFS writes throughout the LUN, causing allocations that do not match the amount of data that is written.

A heavily used NTFS filesystem running without much free capacity eventually uses up all of the capacity unless the filesystem is defragmented periodically. NTFS favors writing into newly allocated space instead of reusing previously written space. NTFS works with thin provisioning initially but can quickly use up more allocation than the amount of data that the filesystem shows as used.

Because thin provisioning uses Controller resources and affects performance, a good use of thin provisioning is for a LUN that has the following characteristics: For example, given 420 GB of file data, the administrator should configure the allocated logical capacity of the LUN to be approximately 470 GB and the addressable logical capacity to be approximately 1 TB.
Note: How much capacity NTFS uses depends on many factors, including the amount of data involved in the write operations, where the write operations are performed, and other factors such as the type of storage used in the storage pool.

On Linux platforms, EXT2 and EXT3 filesystems write metadata over the entire range of logical block addresses (LBAs) of the LUN. The drive is organized into block groups and metadata exists at the beginning of each block group. This configuration typically causes the entire LUN to be provisioned when the administrator creates a filesystem. This full provisioning occurs because the metadata write is below the minimum allocation unit that is used by the Oracle FS Systems. This condition causes the system to expand every allocation extent to the maximum size.

In summary, the success of utilizing thin provisioning depends on the filesystem or the application using the LUN.