According to a recent 3PAR press release, thin provisioning is fast becoming a must-have feature for organizations as they look to procure new enterprise storage systems. It cites a report from TheInfoPro that shows that 87% of surveyed users now consider thin provisioning an important virtualization technology. Much of this increased user interest in thin provisioning is attributable to their growing adoption of server and storage virtualization. However for any organization to realize the full benefits of thin provisioning is dependent upon what an organization’s computing environment looks like when a thinly provisioned storage system is deployed.
Taking advantage of a storage system’s thin provisioning feature is usually less of an issue when thinly provisioned volumes are deployed in conjunction with new applications or servers. Since there is no existing data to migrate or move, these applications or servers will “see” the 100 GB thinly provisioned volume presented to them and only use as much storage space as actually exists. So if the application only needs 10 GB of storage space that is the only space that the storage system that supports thin provisioning will allocate.
Similarly, organizations will realize the benefits of thin provisioning when they use replication software that copies and/or moves data at the file system level from volumes on one storage system to another. When replication occurs at this level, only “real” data is moved in the sense that only the data in the files are moved. So as the file-based replication software copies a file from a “fat’ volume to a new “thin” volume, the only storage that a file consumes on the new storage system’s thinly provisioned volume is the amount of storage that the file need.
But that is where the benefits of a thin provisioning storage system can come to end. Many thin provisioning storage systems are targeted at environments with enterprise applications. In these environments, they use replication software that moves data at the block or volume level, not the file system level. The reasons for moving data at the block level are numerous but the more notable ones are that the application can continue processing while the data migration occurs; the data being moved is in database so a file system migration is not an option; and it may incur less overhead on the server.
The problem that arises with block level migrations is that the replication software is not aware if the block of data that it is moving actually has any data in it. Instead it just blindly copies the block of data from the old storage system to another blog on the thinly provisioned storage system. This is where the problem of this approach shows up. The thinly provisioned storage system interprets this incoming block of data sent over by the replication software as “real” data whether or not the block of data contains any actual data and reserves space for it on the new thinly provisioned volume. The problem that occurs in these circumstances is that the organization realizes no practical benefits from thin provisioning as just as much storage space is consumed on the new thinly provisioned storage system as before.
To overcome this barrier to realizing the benefits of thinly provisioned volumes, two options now exist in the Veritas Storage Foundation suite for users to migrate existing data on existing “fat” volumes to new thinly provisioned volumes. These are:
- Veritas SmartMove. This feature is available to users that have data mounted on a volume that is using either the Veritas File System or Windows NTFS in conjunction with Veritas Volume Manager. Both Veritas File System and Windows NTFs know which blocks are used and unused on the volume. So when Veritas Volume Manager copies data at the block level, it only copies blocks of data with actual data in it to the new thinly provisioned volume on the other storage system. All of this can be done online, without application downtime, and supports any thin storage array. This approach is now available for any Unix/Linux server that is using Veritas Storage Foundation version 5.0 MP3 or Windows 5.1 with NTFS.
- Reclaim before move. The other option that is available on current and prior versions of Veritas Volume Manager is to shrink the file system and volume before the data is moved. This option is, however, not as efficient from a storage perspective as the new SmartMove option. Shrinking leaves some empty blocks behind plus it requires more planning to execute on. However it does contribute to reducing the amount of data that can be migrated online to a thinly provisioned volume be it attached to a Linux, UNIX or Windows server.
How great a role will or should Veritas Volume Manager and Veritas File System play in this rapidly emerging world of thin provisioning? Organizations that are going into networked storage and thin provisioning for the first time or doing data migrations at the file system level may see little initial benefits from Veritas Storage Foundation, at least in conjunction with thin provisioning. It is organizations that support large databases with 60% or lower utilization rates on their current storage volumes that will see the most benefits from a cost and capacity savings perspective.
As TheInfoPro has uncovered in its research, thin provisioning is a feature almost all organizations will encounter as they procure and deploy new storage systems into their environments. But for legacy applications running on “fat” volumes, taking advantage of thin provisioning does not occur automatically. However by using intelligent volume management software from Veritas in conjunction with its file system and replication software, organizations can move into these new thinly provisioned environments and fully realize all of the benefits that thin provisioning delivers now and into the future.