VCF / VCF9 · July 5, 2026

Inside VCF 9.1: A Feature-by-Feature Technical Deep Dive

This post assumes you already know why VCF 9.1 matters (see our overview post) and want the component-level detail — the kind you need to actually plan a design or run a demo.

vSphere Layer

  • vCenter Quick Patch. Rather than patching every RPM regardless of whether it changed, Quick Patch updates only the specific RPMs/binaries with an actual code change in the payload. Net effect: materially faster patch cycles and, for many security fixes, little to no downtime.
  • vCenter resize via API. A single API call plus a reboot resizes vCenter compute/disk — no more manual VM reconfiguration and re-validation. Reachable through the Developer Center API Explorer.
  • Back-in-time support. VCF now formally documents three supported paths to 9.1 — Upgrade, Converge (bring an existing vSphere-based non-VCF deployment into a new fleet), and Import (bring an existing vSphere-based deployment into an existing fleet) — with a defined compatibility table for cases where your source version (e.g., vSphere 8.0 U3j or VCF 5.2.4) was actually released after 9.1 itself.

Storage: vSAN

  • Extended deduplication and compression now spans more cluster types and workload profiles, including — new in 9.1 — dedupe support for data already encrypted at rest, closing a long-standing gap where customers had to choose between encryption and space efficiency.
  • vSAN ESA HCL enforcement is stricter in 9.1 (this is why lab-scale nested deployments increasingly need the HCL bypass workaround — see our dedicated workaround post).

Memory Efficiency

Enhanced NVMe Memory Tiering is arguably the single most consequential efficiency feature in 9.1. The hypervisor actively tiers hot pages to DRAM and cold pages to local NVMe, expanding effective memory capacity without adding physical DRAM, and includes software-based mirroring plus built-in cost-savings analysis so you can quantify the win to a customer’s CFO.

Kubernetes / VKS

  • Scale: a single Supervisor control plane now supports up to 500 clusters, with up to 70% faster provisioning versus prior releases.
  • Multi-network support: cluster nodes can be deployed with multiple vNICs, isolating application, storage, and management traffic — a requirement many regulated customers have been asking for since VKS GA.
  • VM Service Fast Deploy: linked-clone-based deployment spins up workloads (e.g., Ubuntu-based VMs) in seconds rather than minutes.
  • Regional Harbor: a full hands-on-lab-validated pattern for a Regional Harbor instance with replication rules, useful for multi-site container image distribution.

Networking (NSX Inside VCF)

The NSX-specific feature set (Distributed Transit Gateways, VNA clusters, VPC connectivity policies) is large enough that we cover it in a dedicated “What’s New in NSX 9.x” post — but at the VCF level, the notable integration points are:

  • New New-VpcGroup / New-VpcConnectivityPolicy PowerCLI cmdlets to organize VPCs into logical groups with strict, Transit-Gateway-bound communication policies (community, isolated, or custom).
  • New-VpcIpBlock and related cmdlets to define IP blocks backing subnets, edge services, or external IPs directly through automation.
  • EVPN-based interoperability with Arista Networks fabrics, extending NSX overlay networking into physical EVPN/VXLAN environments without a translation layer.

Platform Engineering & Operations

  • VCF Operations 9.1 introduces a workflow for creating a workload domain without a vSphere cluster — vCenter and NSX Manager get deployed/joined first, with the cluster added later. Important caveat: you cannot patch or upgrade components in that domain, or any domain sharing its NSX Manager, until a vSphere cluster is added.
  • Security Posture Management assesses the environment against industry benchmarks (PCI-DSS, NIST, CIS, and the new “VCF 9 General Controls 1.0” benchmark) directly from VCF Operations, with additional compliance packs available from the Broadcom VCF Solutions Catalog.
  • Configuration Management (formerly “Configuration Drift”) now supports templates for vSphere Configuration Profile (VCP)-enabled clusters.
  • Upgrade sequence change: component upgrade order shifted to NSX Manager → vCenter/ESX (NSX host bits are now bundled with the ESX upgrade) → NSX Edge clusters last. This is a change from pre-9.x sequencing and worth calling out explicitly in change-management documentation, since runbooks written against older VCF versions will have the wrong order.

Automation & Cloud Management

  • VCF Automation (VCFA) 9.1 upgrades via a side-by-side migration rather than in-place patching: Fleet Lifecycle stands up a new 9.1 cluster alongside the existing 9.0.x nodes, migrates data, then decommissions the legacy cluster.
  • Multiple external connections per organization — tenants can now have more than one exit point for outbound traffic, supporting both centralized (Tier-0/Tier-0 VRF) and distributed VLAN-backed connections, configured through the Provider Management UI.
  • Non-disruptive VM import via the new Mobility Operator (a core Supervisor service) allows running VMs to move into VCFA-managed namespaces without reboot or re-IP — a major improvement over the 9.0 “Change management layer” workflow, which required both.

Practical Design Takeaway

Almost every feature above reduces either (a) the operational burden of running VCF at scale, or (b) the hardware footprint required to run it. When you’re building the business case for a customer, that’s the frame to use: 9.1 isn’t a checkbox upgrade, it’s a cost-and-complexity reduction release layered on top of the 9.0 architectural shift.

You may also like...