Dell Unity: ESXi Datastore Reports Larger Size Than The Coordinating Unity LUN

VMware/Broadcom ESXi datastore shows higher used space when Data Reduction is enabled on the Unity array. 

Customers might have queries such as:
1. Why is the space utilization of the Datastore different on the Unity vs. the ESXi side?
2. Is Data Reduction (compression) working properly? If so, then why are there no savings reported on the host-side?
3. What is the purpose of this Data Reduction if the hosts do not see any savings on their end. 

 VMware vCenter is not aware of any Data Reduction/space saving efforts occurring on the Unity. These are Unity features - not host features - so the host has no insight.

Example:

• A VMFS datastore named "LI_UNITY400" is created on Unity with Data Reduction enabled. 

          Total size: 2 TB
          Allocation size: 570 GB
          Current savings: 700 GB

• On the VMware Datastore "LI_UNITY400" shows the size used space of 1.27 TB which is more than 570 GB Allocation size on Unity.
• If you add up the "Allocated size" and "Current savings"  on Unity (570+700 GB= 1.27 TB) , it gives you the same "used space" as on ESXI host because the host is not aware of the Data Reduction Feature on Unity.

If the data is compressed on the Unity array, it must first be uncompressed before the data is sent to the host.  If the compressed data already resides in System Cache, the data is uncompressed to a temporary location before being sent to the host, and then the temporary location is released. If the requested, compressed data resides on disk, the data is first read into System Cache before being uncompressed to a temporary location, and the host is sent the data.

Data is never uncompressed on disk due to the read operation, as this would reduce the amount of savings on the storage resource. 

Dell Unity: Data Reduction

The array is functioning as designed.

NOTE:  In Dell Unity OE version 4.3 and later, Dell Unity Data Reduction replaces Dell Unity Compression. Data reduction includes compression and deduplication logic within the space savings algorithm.