Technical white paper Page 7
• Environments that require host encrypted volumes—Writing blocks of zeros to a host-encrypted volume on a newly created HPE 3PAR
StoreServ thin-provisioned volume will cause space to be allocated on the TPVV because the encryption alters the content of the blocks.
Applying encryption to thin-provisioned volumes that already contain data or rekeying them also inflates the zero blocks, making the
volume consume space as if it was full-provisioned. Attempting to rethin the volume by writing zeros to allocated but unused space is not
possible as well. As a result, host encryption and thin provisioning do not cooperate well.
• Environments that require SAN encrypted volumes—Like host-based encryption, encryption by a device in the data path (for example,
SAN switch) will also alter the data stream so that blocks of zeros written by the host are not passed onto the storage. A notable
exception is Brocade SAN switches. With the introduction of Fabric OS 7.1.0, the Fabric OS encryption switch can automatically detect if a
disk LUN is a thin-provisioned LUN. If a LUN is detected as being thin-provisioned, then first-time encryption and rekey are done on the
allocated blocks only. This thin-provision LUN support requires no action by the user.
• Copy-on-write file systems with low data reduction ratios—File systems that write to new blocks rather than overwrite existing data are
not suitable for thin provisioning, as every write will allocate new storage until the volume is fully allocated. An example of a copy-on-write
(CoW) file system is Oracle Solaris ZFS.
Thin
The use of data reduction technologies has the significant operational benefit of reducing storage consumption. However, there are certain
scenarios where data reduction may not be of benefit and regular thin provisioning can be a better choice such as
• The data is to be stored solely on HDDs or tiered to HDDs using HPE 3PAR Adaptive Optimization.
• Environments that use application or file-system-based encryption, deduplication, or compression.
• High write workloads—With thin provisioning, metadata is only updated when space is allocated and subsequent overwrites do not
generate any metadata updates. Therefore, intensive write workloads can achieve higher performance without data reduction.
Compression
Compression is ideal for data that does not have a high level of redundancy. Data sets that are good candidates for compression include
• Databases—Most databases do not contain redundant data blocks but do have redundant data within blocks so they can benefit from
compression.
• Virtual machine (VM) images—VMs where the application data size far exceeds the operating system binaries size may not yield
significant deduplication savings but can benefit from compression of the application data.
• Virtual desktop infrastructure (VDI)—Client virtualization environments with hosted nonpersistent desktops can achieve excellent
compression ratios.
Data with a low compression level should be stored on thin-provisioned volumes. Data sets that are not good candidates for compression
include
• Compressed data—The use of application or host-based compression will create a stream of unique data that will not benefit from
storage compression.
• Encrypted data—The use of host or SAN encryption will also result in a stream of unique data that will not benefit from storage
compression.
Use the Compression estimate tool to check the compression ratio of existing volumes before conversion to compression volumes.
Deduplication
Deduplication is ideal for data that has a high level of redundancy. Data sets that are good candidates for deduplication include
• Virtual machine (VM) images—The operating system binaries from multiple VMs can be reduced to a single copy by deduplication. Note
that the application data within the VMs may be unique will therefore not benefit from storage deduplication but may benefit from
compression.
• Virtual desktop infrastructure (VDI)—Client virtualization environments with hosted persistent desktops can achieve excellent
deduplication ratios. Note that nonpersistent desktops using linked clones will achieve lower ratios as the desktops share a single golden
image instead of each desktop having its own operating system image.
• Home directory and file shares—Users often store copies of the same file in their private workspaces and therefore storage deduplication
can offer significant space savings.