EMC XtremIO and Pure Storage capacity savings technologies
EMC XtremIO and Pure Storage arrays use capacity-savings technologies including:
- De-duplication - removing redundant data.
- Compression - making the object smaller using standard compression algorithms.
- Thin provisioning - using virtualization to give the appearance of having more physical resources than are actually available.
As an example, let us look at the LUN creation process for these arrays. Suppose we were to provision a logical object (LUN) on an EMC XtremIO or Pure Storage array, and wanted to make it 100 GB.
After creating the LUN, we can inspect the LUN Details view to see that the Total Size property shows 100 GB, the size of the LUN required.
We can investigate the Consumed Capacity value. Immediately after the LUN is created this value will be a very small fraction of the 100 GB. It will, of course, grow as the user adds data to the LUN. The value displayed should be exactly the net amount of data which has been sent to the LUN. Since this is a logical value, as seen by the host computer operating system, it cannot be greater than the Total Size.
If we then look at the Storage Pool Details view we should see the Total Subscribed Capacity has increased by 100 GB.
Depending on the array total capacity and the Total Subscribed capacity, the Oversubscribed Capacity can stay at 0, or grow to a value between 0 and 100 GB. This value shows how much the sum of the LUNs Total Size values for the pool is bigger than the Total Usable Capacity.
Over the lifetime of the LUN, Total Capacity of the LUN and its share in the Total Subscribed Capacity for the array will not change, unless the storage administrator expands or shrinks the LUN.
In most cases the Consumed Capacity of the LUN will not decrease, with the exception for LUNs owned by certain operating systems. For example, if VMware is used, VAAI is able to communicate to the array a list of unused blocks of data - these blocks will be removed, decreasing the Consumed Capacity.
The influence of adding the 100 GB LUN on the Usable Capacity for the array cannot be predicted with 100% certainty. Because of global de-duplication, adding the data can leave the free usable capacity unchanged in extreme situations. Therefore, any prediction for run-out time is statistically valid at the whole array level.