VCF 9.1 Upgrade Benefits: Unlocking vSAN ESA Auto-RAID for Smarter Storage

Auto RAID capabilities in VMware Cloud Foundation 9.1 represent an important evolution in how storage efficiency and resilience are managed within modern virtual infrastructure. As we continue reviewing the benefits and features of the VCF9.1 upgrade, this blog and the VCF 9.1 Auto RAID capabilities it references apply specifically to the vSAN Express Storage Architecture (ESA) within VMware Cloud Foundation 9.1. These capabilities are not supported on the vSAN Original Storage Architecture (OSA), which is also available for use with VCF 9.1.

For readers not familiar with the architectural differences between vSAN ESA and OSA, we encourage watching podcast Example Architectural Decisions - VMware vSAN Architecture (ESA vs OSA).

Let’s dive into the enhancements to vSAN ESA Auto RAID in VMware Cloud Foundation (VCF) 9.1, including its underlying logic, resiliency improvements, effective capacity gains, and the new single policy operational model.

vSAN Express Storage Architecture (ESA) Resiliency

One of the major architectural advancements introduced with vSAN Express Storage Architecture (ESA) in vSphere 8 was the dramatic improvement in the performance and efficiency of Erasure Coding (RAID-5/RAID-6). Historically, many architects/customers preferred RAID-1 mirroring due to the performance overhead associated with traditional erasure coding in OSA environments. 

ESA significantly reduced those overheads through a redesigned storage architecture optimised for modern NVMe flash, high-core CPUs, and high-speed networking. As a result, RAID-5 and RAID-6 have become not only a viable, but highly attractive design choices since ESA was introduced due to its excellent performance and capacity efficiencies.

VCF 9.1 Auto-RAID builds on these vSAN ESA improvements by dynamically selecting the optimal resilience scheme automatically, making Auto-RAID the recommended default choice for most new VCF storage deployments. This greatly simplifies storage policy design and operational management.

For additional architectural guidance, readers are encouraged to watch the podcast On-line with an Architect: Example Architectural Decision – vSAN Storage Policy.

What is Auto-RAID?

Auto-RAID is a vSAN ESA capability introduced in VCF 9.1 that automatically manages storage resilience and RAID selection based on cluster topology and operational conditions. Instead of administrators manually selecting RAID-1, RAID-5, or RAID-6 policies, Auto-RAID dynamically determines the most efficient and resilient layout.

Auto-RAID Functionality

• Auto-RAID is available only with vSAN ESA and not with traditional OSA.
• Auto-RAID prioritises erasure coding because since the release of vSAN ESA, Erasure Coding significantly reduces historical parity overhead penalties while providing greater usable capacity than mirroring (i.e.: FTT1 being Two-Way & FTT2 being Three Way Mirroring)
• Auto-RAID continuously evaluates cluster topology and automatically adjusts object placement.
• Resilience decisions are object-based and policy-driven rather than globally disruptive.
• Transitions occur online with minimal operational impact.

Single Universal Policy Model

VCF 9.1 introduces a simplified single-policy model for Auto-RAID. Rather than maintaining multiple storage policies for different cluster sizes and RAID types, administrators can apply one Auto-RAID policy and allow ESA to optimise object placement automatically as the cluster evolves.

Auto-Raid Logic in VCF 9.1

Detailed Auto-RAID Logic

One of the most important architectural improvements in VCF 9.1 Auto-RAID is the simplified and highly consistent resilience logic used by vSAN ESA. Unlike previous versions of vSAN, especially OSA, where administrators needed to carefully select and manage multiple storage policies manually, Auto-RAID now dynamically applies the most appropriate resilience scheme automatically based on cluster topology.

The logic used by Auto-RAID is intentionally optimised around the enhanced erasure coding performance delivered by vSAN ESA. Whenever possible, ESA will prefer space-efficient erasure coding (RAID-5 or RAID-6) instead of traditional RAID-1 mirroring. The only exceptions are site resilience for stretched clusters and host resilience for 2-node deployments, where mirroring is still required due to cluster size constraints.

This design dramatically simplifies storage policy management while delivering excellent capacity efficiency and strong resiliency characteristics.

Effective Capacity Concepts

VCF 9.1 introduces improved effective capacity reporting using standardised Auto-RAID overhead calculations. This allows storage consumption reporting to more closely resemble traditional enterprise storage platforms.

When Auto-RAID Re-evaluates RAID Placement

• Hosts are added or removed
• Cluster topology changes
• Maintenance mode operations complete
• Storage policies change
• Capacity rebalance operations occur
• Object repair or resynchronisation events occur

Operational Advantages

• Simplifies storage policy management
• Reduces human error
• Improves storage efficiency automatically
• Enables incremental cluster growth
• Aligns with self-optimising private cloud infrastructure

Supported RAID Stripe Widths and Parity Layouts

vSAN ESA in VCF 9.1 supports multiple RAID stripe widths and parity schemes depending on the selected resilience level, cluster size, and topology. Auto-RAID automatically selects the most appropriate layout based on available fault domains and resiliency requirements.

One of the major improvements introduced with ESA is the significantly enhanced performance and efficiency of erasure coding, making RAID-5 and RAID-6 practical for many more production workloads compared to historical OSA-based designs.

Architectural Recommendations: Cluster Sizing

One of the most important architectural design decisions when deploying vSAN including with ESA Auto-RAID in VCF 9.1 is determining the appropriate cluster size to maintain the desired level of resiliency during host failures, maintenance operations, or unexpected outages. Although RAID-6 (4+2) becomes available once a cluster reaches six hosts, architects should carefully consider what happens when hosts become unavailable. 

A six-host cluster is technically capable of supporting RAID-6, but the loss of a single host reduces the available fault domains below the minimum required to maintain the 4+2 stripe layout. If the outage persists beyond the default 60 min repair timer, Auto-RAID may dynamically restripe objects down to RAID-5 (2+1), reducing the cluster from FTT=2 to FTT=1 resiliency which also adds a non-trivial overhead to the cluster during this operation.

For environments that specifically want to benefit from the stronger resiliency characteristics of RAID-6 and maintain the 4+2 stripe during host outages or maintenance operations, the recommended design approach is to deploy clusters with at least seven hosts, while eight hosts is the preferred architectural target.

Stretched Clusters

One important architectural detail is that when Auto-RAID selects FTT=1 using RAID-5, it will always use the 2+1 RAID-5 stripe width. The 4+1 stripe width configuration available in is not used by Auto-RAID.

Important Notes

• Auto-RAID in VCF 9.1 always prefers erasure coding when sufficient hosts and fault domains are available.
• Auto-RAID uses the modern 2+1 RAID-5 scheme for FTT=1 rather than the older 4+1 option.
• RAID-6 in ESA uses a 4+2 dual parity scheme for improved resiliency and storage efficiency.
• Stretched clusters combine site-level RAID-1 mirroring with local RAID-5 or RAID-6 protection inside each site.
• 2-node clusters continue to rely on RAID-1 style mirroring for site/host resilience.

Recommended Cluster Sizing

This guidance provides practical cluster sizing recommendations for both RAID-5 (FTT=1) and RAID-6 (FTT=2) Auto-RAID deployments. While Auto-RAID dynamically adjusts resiliency based on available hosts, architects should carefully consider how cluster sizing impacts critical factors such as usable capacity, maintenance operations and unexpected host outages.

From a practical operational perspective, four-host clusters provide the most balanced architecture for environments adopting Auto-RAID with RAID-5 resiliency & eight-host clusters recommended for environments adopting Auto-RAID with RAID-6 resiliency.  These cluster sizes allow the environments to maintain the desired resiliency (RAID) level/s during maintenance operations, rolling upgrades, transient outages & failed host/s. These recommendations become increasingly important for mission-critical workloads and large-scale private cloud environments where maintaining consistent performance and resiliency levels during maintenance/failure events is essential.

Can RAID-5 with a 4+1 stripe still be used?

Yes RAID-5 with a 4+1 strip can still be used, but not with Auto-RAID. For environments with 6 hosts which may be capacity constrained, manually configuring a storage policy with RAID 5 (4+1 stripe) will provide the highest level of usable capacity and N+1 resiliency ensuring the failure (transient or permanent) can be tolerated by the cluster. This does however slightly increase the complexity within the environment that Auto-RAID aims to avoid.

Strategic Importance

Auto-RAID represents a major architectural evolution in VMware storage management. Rather than administrators manually designing resilience layouts, ESA dynamically optimises protection schemes automatically based on cluster state and topology. This aligns closely with the broader VCF 9.1 strategy around automation, self-optimising infrastructure, and simplified private cloud operations.

Stay tuned for an upcoming instalment where we will discuss vSAN Data efficiency enhancements in VCF 9.1, or catch up on how to improve efficiency with memory tiering.

And if you’d like to discover how Auto RAID can transform your storage strategy and simplify your VMware environment, get in touch with Zaleo Consulting. Our team of specialists can help you leverage ESA’s intelligent protection models, align your architecture with VCF 9.1 best practices, and create a self optimising infrastructure that reduces complexity while enhancing resilience so you can run a smarter, more efficient private cloud with confidence.