Amazon Elastic VMware Service (Amazon EVS), the AWS service that you can use to deploy and run your on-premises VMware Cloud Foundation (VCF) stack directly on AWS, now supports the latest VCF 9.1 release. For organizations planning their next VMware upgrade or VMware modernization, this announcement introduces new opportunities to improve infrastructure efficiency and reduce costs.
From a cost-efficiency perspective, two of the most impactful features introduced in VCF 9.1 are:
- Enhanced Memory Tiering, which doubles the effective memory so you can add twice as many VM hosts in the same cluster.
- Disaggregated storage that allows storage services other than virtual storage area network (vSAN), including FSx for ONTAP, to serve as the primary storage. This way, compute and storage can scale independently and overprovisioning is reduced.
Together, these capabilities create a new architectural blueprint that can significantly reduce the infrastructure footprint and cost of VMware workloads on AWS.
In this post, we explore how enhanced Memory Tiering in VCF 9.1, combined with Amazon FSx for NetApp ONTAP (FSx for ONTAP) as the primary storage on Amazon EVS, can help organizations improve infrastructure efficiency, increase VM density, and significantly reduce costs while modernizing VMware environments.
How Memory Tiering lowers total cost of ownership (TCO) in VCF 9.1
VCF 9.1 introduces enhanced Memory Tiering, a feature that can double a host’s effective memory capacity, improving CPU utilization. It does so by using NVMe devices as a secondary memory tier alongside DRAM, moving infrequently accessed (cold) memory pages to NVMe while leaving active (hot) pages in the faster DRAM tier.
When using the default 1:1 ratio (equal amounts of DRAM and NVMe), Memory Tiering effectively doubles the memory available to a host, leading to a 200% increase in VM density for suitable workloads. With more VMs running on each host, organizations might also be able to reduce the number of hosts they need. VMware estimates that this feature alone can result in up to 40% lower TCO.
Memory Tiering is best suited to workloads with natural hot/cold memory patterns, such as general virtualization, virtual desktop infrastructure (VDI), and DevTest environments. According to Broadcom, clusters with less than 50% active memory and ample free CPU capacity are strong candidates to benefit from this feature. Conversely, they advise against using Memory Tiering for latency-sensitive workloads or workloads with large active memory footprints.
FSx for ONTAP: Storage efficiency and data management
VCF 9.x now officially supports non-vSAN storage options, such as FSx for ONTAP, as primary storage, enabling a disaggregated storage design that separates compute from storage. Organizations can now scale capacity independently, reducing overprovisioning, better aligning infrastructure to workload needs, and saving on costs.
FSx for ONTAP brings the same efficiency-driven approach as Memory Tiering to the storage layer. It automatically tiers infrequently accessed data to lower-cost capacity pools while keeping active data on high-performance SSDs. Added to this are the NetApp® ONTAP® core storage efficiencies—thin provisioning, deduplication, and compression—that further reduce your storage footprint. Space-efficient NetApp Snapshot™ copies and thin clones further reduce storage consumption.
Beyond its cost benefits, FSx for ONTAP provides enterprise-grade ONTAP data management capabilities. These include integrated data protection that supports demanding recovery point objectives (RPO) and recovery time objectives (RTO), automatic multi-Availability Zone (AZ) failover with zero rehydration delay, seamless data mobility with SnapMirror® unified replication, and familiar operational workflows for storage teams already using ONTAP on premises.
Using FSx for ONTAP as the primary storage offers further advantages in an EVS architecture:
- Freeing all NVMe resources for Memory Tiering: In a vSAN architecture, local NVMe devices support the shared storage layer. By moving primary storage to FSx for ONTAP, those NVMe resources can be dedicated to VCF 9.1 Memory Tiering instead of being shared with storage services, helping to maximize benefits.
- Supporting a lean two-host starting point: With FSx for ONTAP, Amazon EVS can run on a smaller configuration, lowering the entry point for VMware modernization on AWS. A cluster with vSAN as the primary storage needs a third host solely to hold a quorum vote in case of conflicting data after a failure. FSx for ONTAP removes this requirement, because data protection is handled in the storage layer—the file servers run as a highly available pair with automatic failover—so host count is governed only by vSphere HA restart capacity. We’ll discuss this capability in detail in the next section.
The following diagram shows a traditional EVS deployment using vSAN with a 4 host minimum on VCF 5.x
VCF 9.1: Amazon EVS with FSx for ONTAP ( 2 host deployment)
Starting leaner with a two-host architecture using FSx for ONTAP
A standard cluster using vSAN requires a third witness server (host) to break ties and prevent data conflicts during a crash. As explained above, using FSx for ONTAP as the primary storage bypasses this three-host minimum requirement, providing a lower-cost entry point to Amazon EVS. This two-host architecture consists of two EVS hosts, an NFS datastore on FSx for ONTAP for the management domain, and a single VMware NSX Manager. In Amazon EVS, the two-host management domain uses NFS storage.
This lean configuration provides a lower-cost entry point for VMware modernization, suitable for several use cases:
- Disaster recovery (DR) pilot-light deployment: Organizations can replace a costly 1:1 secondary data center with a two-host pilot-light environment. They use SnapMirror to replicate data from ONTAP on premises to FSx for ONTAP, and scale EVS compute during a disaster with zero data rehydration delay.
- Proof of concept and phased migrations: For organizations starting their cloud journey, a two-host cluster can be used as a low-cost environment for validation and early migration planning.
- DevTest environments: The smaller footprint is a natural fit for development and testing environments. Many testing environments don't require large production-scale clusters. A two-host deployment combined with FSx for ONTAP’s thin cloning capabilities can provide an efficient development environment.
- Small VMware footprints: Small projects can move to Amazon EVS cost effectively, with fewer hosts.
It’s important to note that the two-host architecture introduces tradeoffs. The VCF management plane consumes a meaningful share of cluster CPU and memory, leaving less headroom for production workloads. If a host fails, vSphere HA can restart workloads on the remaining host, but recovery time may be higher because capacity is limited. If the host running the single NSX Manager fails, existing workloads can continue running and FSx for ONTAP continues serving data, but network configuration changes are temporarily unavailable. It’s therefore essential to ensure proper backup of the NSX Manager.
Whether the design is two-host, three-host, or a larger cluster, each option has valid use cases. The right choice depends on workload requirements, resilience objectives, growth expectations, and operational priorities.
The case for running VCF 9.x on Amazon EVS
VCF 5.2 reaches End of General Support on October 2027, making lifecycle planning a near-term priority for VMware customers.
At the same time, infrastructure refreshes are becoming costlier and harder to plan as manufacturers face extended lead times. Amazon EVS offers an alternative path: Run VCF 9.x on AWS, preserve the familiar VMware operating model, and create a cloud foundation for future modernization.
Summary
Support for VCF 9.x on Amazon EVS, combined with the architectural shift it enables, offers a powerful path to modernizing VMware workloads.
- With VCF 9.1 Memory Tiering, you can double the effective memory per host, increase VM density by up to 200%, and reduce TCO by up to 40%.
- With FSx for ONTAP, you gain substantial additional savings through its built-in cost-performance capabilities. It also frees all NVMe resources for use with Memory Tiering
, and provides a multitude of leading enterprise storage capabilities.
For organizations evaluating their next VMware move, the conversation is no longer limited to upgrading software. It's an opportunity to rethink infrastructure design, improve efficiency, and potentially lower costs at the same time.
To learn more read these blogs:
Using Amazon Elastic VMware Service with FSx for ONTAP to optimize costs and modernize VMware
Increase your cluster density with Memory Tiering in Amazon Elastic VMware Service