What are the key differences among various types of storage thin provisioning implementations?
The key differences among storage thin provisioning technologies relate to the level of granularity with which each technology can apply itself and how well it handles thin provisioning after new thin volumes have been created.
When creating new volumes, traditional thick provisioning allocates more space than is needed to simplify future expansion. The typical use case for thick provisioning is to support databases or specific applications. This practice can also make manual storage volume expansion easier.
Thin provisioning logically reserves enough space when a volume is created to support all expected future expansion, but physically allocates that space closer to the time it's written to, and in smaller blocks, so less physical storage is committed and sitting idle. This controlled overallocation must be carefully managed so that physical capacity is made available before it's needed.
When setting up a new storage system with the ability to thinly provision storage, users often want to populate it with previously provisioned thick volumes. This requires the new storage system to make those thick volumes thin. This is accomplished by the storage system first detecting the zeros written as placeholders to the original thick volumes and then not writing them to the thin volume.
When data is deleted from thin volumes, it can sit idle, making that volume fatter. To address this, some systems can communicate with the storage volume manager, application or file system to identify the capacity freed by these delete activities and make it available.
There are some other nuances to thin provisioning, but generally, the storage system will include thin provisioning as a feature that is sometimes optional. When deciding whether to use this feature, consider the expected ratio of storage saved by each storage system (based on the data types you will store on it and the activity it will see), the cost of the thin provisioning feature and whether it can be configured at the volume level.
About the expert
Eric Slack is an analyst at Storage Switzerland, an IT analyst firm based in Fort Worth, Texas, which focuses on the storage and virtualization segments. Eric has more than 20 years of experience in high-technology industries.
This was first published in May 2013