vSAN space efficiency: deduplication, compression and erasure coding

vSAN space efficiency features reduce the raw capacity needed to store and protect data. They differ between ESA and OSA, and some interact with policy settings and encryption.

Technique	vSAN OSA	vSAN ESA
Deduplication and compression	All-flash only; cluster-wide setting applied per disk group; dedupe within a disk group, not across; about 5% raw capacity overhead	Deduplication not available
Compression only	Cluster-wide, per disk; suits workloads that do not dedupe well, such as OLTP	Default on, per object via storage policy; can be turned off for selected VMs with a custom storage policy; applies to new writes only
RAID-5/6 erasure coding	All-flash; RAID-5 from 4 fault domains, RAID-6 from 6	RAID-5 from 3 hosts (2+1) or 4+1 at 6 or more; RAID-6 from 6
TRIM/UNMAP	Enable Guest Trim/Unmap in vSAN advanced options, then power-cycle VMs; VM hardware version 13 or later	Enabled by default
Thin provisioning	Default (object space reservation 0)	Default

Further points:

Enabling or disabling deduplication and compression triggers a rolling reformat of every disk group or storage pool, which can take a long time, so it should not be done often. Before disabling, check there is enough physical capacity.
With deduplication and compression on an OSA cluster, policies must use 0% or 100% object space reservation; thick provisioning makes deduplication and compression savings ineffective. Disks should be added as whole disk groups, a single disk cannot be removed from a disk group, and one disk failure fails the whole disk group.
VM encryption encrypts data before it reaches storage, so deduplication and compression may not be effective for encrypted VMs. vSAN data-at-rest encryption (which needs a key provider) is applied after compression on ESA and after deduplication on OSA; data-in-transit encryption needs no key server.
The dedupe and compression ratio is Used Before divided by Used After in the vSAN Capacity monitor.
Guest UNMAP works offline (Linux fstrim) or inline (mount -o discard; NTFS performs inline unmap by default).

Common trap: Selecting thick provisioning (100% object space reservation) together with deduplication and compression and expecting savings - full reservation makes deduplication and compression savings ineffective.

Get the whole book

This note is one section of Ultra Transcenders 2V0-17.25: VMware Cloud Foundation Administrator, an independent study guide that explains every topic the exam covers by technology, with comparison tables, diagrams and the common traps, plus a glossary linked to Broadcom TechDocs.

Amazon.co.ukKindle: coming soonPaperback: coming soon

Amazon.comKindle: coming soonPaperback: coming soon

Publishing soon on Amazon in Kindle and paperback editions.

About the book · 2V0-17.25 terms in the glossary · All 2V0-17.25 study notes

More 2V0-17.25 study notes

Components renamed or replaced in VMware Cloud Foundation 9.0
Aria, Cloud Builder, ESXi and vSphere with Tanzu: what each is called in VCF 9.0 and what replaced it.
vSAN ESA vs vSAN OSA in VCF 9.0
How the Express and Original Storage Architectures differ, and when each fits.
NSX VPCs: centralized vs distributed connectivity
Tier-0 on an Edge cluster or a distributed transit gateway: what each gives VPCs and what it rules out.
VCF Operations metrics, properties and super metrics
The difference between metrics and properties, and when a super metric is the answer.
VCF Operations policies: priority, inheritance and what they control
How VCF Operations policies decide what is collected and alerted on, and which policy applies to an object.
Moving from Aria Operations for Logs 8.x to VCF Operations for logs
Log forwarding vs Log Data Transfer, and what moves to VCF 9.0.
VCF Automation multi-organization tenancy
All Apps and VM Apps organizations, and how providers carve up a VCF Automation instance.