Impact of turning deduplication and compression off on a large production VM.

Products

• VMware vSAN
• VMware vSAN 8.x

Issue/Introduction

SQL VMs running on a vSAN 8.x OSA cluster configured with the deduplication and compression feature were experiencing performance issues. 
It was suggested to disable the deduplication and compression feature to improve the VM performance, and this KB explains what impact, if any, there will be by doing so.

Environment

VMware vSAN 8.x

Resolution

Overview:
Disabling the deduplication and compression feature in vSAN requires changing the disk format on each disk group within the cluster. This process involves evacuating data from the disk group, removing the disk group, and recreating it without support for deduplication and compression.

1. Expected Performance Impact on the Cluster:
 During this operation, you should expect some performance degradation due to additional resynchronization (resync) activities required to move and rebuild data.
 The impact may vary depending on cluster size and workload but is generally noticeable during the disk group recreation phase.

2. Expected Performance Impact on Virtual Machines (VMs):
 VMs running on the cluster may experience latency or reduced performance, especially those that are sensitive to latency.
 Large VMs, such as one sized around 31 TB, could be more affected depending on their application workload characteristics.
 It is difficult to provide exact performance hit numbers because they depend heavily on specific workloads and environmental conditions.

3. Estimated Time for Completion:
 The duration of this operation depends primarily on:
 The number of hosts in the cluster
 The total amount of data stored on the disk groups being reformatted
 You can monitor progress through the Tasks and Events tab in your management interface to track completion status.

 

Additional Information

Recommendations:
 Schedule this operation during periods of low activity to minimize impact on production workloads.
 Consider backing up critical data before starting the process as a precaution.
 Monitor system performance closely throughout the operation to identify any issues early.