Tools / Guides / VCF Reference Architecture Builder
// Guide · Design & architecture

VCF Reference Architecture Builder

A visual VCF topology designer covering logical, physical, and storage layers in one workspace. Start from a template (PoC, single domain, full mgmt + WLD, stretched cluster), customise the layout, validate against VCF 9 best practices, then export SVG diagrams, bill of materials, network plan, and vendor-specific switch configurations — all from a drag-and-drop interface.

VCF 9.xDrag & Drop6 templatesStretched ClusterCisco · Arista · JuniperNSX BGP PeeringSVG · BOM Export
Open the tool Jump to walkthrough

Quick start

  1. Pick a template — VCF Proof of Concept, Standard Management Domain, Full Standard (Mgmt + WLD), Stretched Cluster, or start blank.
  2. Switch between layers — Logical (clusters, networks), Physical (racks, hosts, switches), Storage (vSAN/NFS/FC).
  3. Drag & drop nodes to customise — add hosts, racks, switches, edge clusters; resize, label, and connect them.
  4. Validate — the live validator flags VCF 9 best-practice violations (cluster size, HA, network redundancy, storage policy).
  5. Export — SVG diagram, BOM list, IP plan, switch CLI config (Cisco NX-OS, Arista EOS, Juniper).
On this page

When to use this tool

Use this tool when you need to:

💡
Best for greenfield design The tool excels at greenfield topology design. For brownfield (modifying an existing VCF), use it to document your target state, then plan migration.

How it works

VCF deployments span three views of the same topology:

The builder lets you author each layer independently but shares state between them — change host count in physical, the logical cluster updates accordingly. The validator runs continuously across all three layers.

Templates ship the standard VCF 9 reference designs as starting points; you customise from there rather than building from blank.

Step-by-step walkthrough

1. Pick a template

Templates available:

Pick the closest match to your target. You can switch templates later but it resets the canvas.

2. Logical layer — clusters and networks

Click Logical tab. You see boxes for each workload domain, the management domain, and the NSX edge cluster. Within each: hosts, network plane CIDRs (mgmt, vMotion, vSAN, TEP, edge uplinks), and any vSAN-specific config. Drag boxes to rearrange, click any element to edit its properties (host count, CIDRs, VLAN IDs).

3. Physical layer — racks, hosts, switches

Click Physical tab. You see racks (one per row), with hosts and ToR switches inside each rack. Drag hosts between racks to rebalance. Add new racks for higher host counts. Each rack shows: rack name, host count, ToR switch model, power feed, U-units used.

4. Storage layer — vSAN/NFS/FC

Click Storage tab. Pick architecture per cluster:

For vSAN, set FTT policy and disk count. The builder calculates usable capacity automatically (uses the same math as the vSAN Capacity Calculator).

5. Validation panel

The validator runs continuously and shows results in the side panel. Common findings:

Findings are colour-coded: red (must fix), amber (should fix), info (FYI).

6. Choose vendor for switch config

Pick from Cisco NX-OS (Nexus), Arista EOS, or Juniper Junos. The exported switch config will use that vendor's syntax.

7. Export

Multiple export formats:

Examples

Example · Standard Management Domain (production starting point)

Template: Standard Management Domain. Customisations:

  • 4 hosts, 64 cores each, 1 TB RAM, 8× 3.84 TB NVMe (vSAN ESA)
  • 2 racks, 2 hosts per rack, 2× Cisco Nexus 9300 ToR per rack
  • FTT=1 RAID-1, 30% slack space
  • 3-node NSX Manager cluster + VIP
  • 2-node NSX Edge cluster on dedicated edge rack

Validator: ✓ all green. BOM: 4 hosts × Dell R750, 4× Nexus 9300, 2× edge appliances, ~32 NVMe drives.

Example · Stretched cluster across two sites

Template: Stretched Cluster (2-site + Witness). Customisations:

  • 3 hosts per site (6 hosts total), witness appliance at third site
  • FTT=1 with site-mirror policy (data on both sites)
  • L3 inter-site link with low latency requirement
  • Single management domain spanning sites

Validator: amber — recommend ≥4 hosts per site for production. Capacity calculator: half the total raw is usable due to site mirroring.

Common mistakes

Forgetting to switch layers when designing A common mistake is designing only on the logical layer and missing physical constraints (rack U-units, power, cooling). Always cycle through all three layers before exporting.
Over-collapsing into one rack Single-rack designs save space but eliminate rack-failure resilience. The validator flags this. Production deployments should span at least 2 racks for any cluster requiring HA.
Edge cluster co-located with management NSX edge clusters generate north-south traffic and need different network connectivity (BGP to physical fabric). Co-locating them with management in the same rack creates routing complications. Best practice: separate edge rack.
Picking template, then completely rebuilding If you find yourself redoing 80% of a template, you picked the wrong template. Switch templates from the menu — it resets the canvas to the new template's defaults rather than fighting an unsuitable starting point.
Switch config without BGP route-maps Exported BGP config peers with NSX edges but doesn't include route-map filters by default. Add filters before applying — accept only legitimate workload CIDRs from NSX, and don't leak management to NSX.

Tools that pair well with VCF Reference Architecture Builder:

FAQ

Can I save my design and come back later?
Yes — the design persists in browser local storage automatically. Export to JSON to back up to disk, then re-import later or share with a teammate.
Does the SVG export include all three layers?
You export each layer separately. For documentation, export logical + physical + storage as three SVGs and combine in your design doc.
Why does the validator complain about my edge cluster?
The most common edge cluster violations: (1) edges in same rack as management (no rack-failure tolerance for routing), (2) only 1 edge node (no HA), (3) edge cluster spans too many racks (latency between edges), (4) BGP peering not configured.
Can I model multi-WLD topologies?
Yes — start from "Full Standard (Mgmt + WLD)" template and add additional workload domains using the "Add Workload Domain" button on the logical layer.
Does the BOM include actual prices?
No — BOM is component counts and SKU suggestions. Pricing varies by vendor, region, and contract. Get quotes from your reseller using the BOM as the requirements list.
Can I import an existing VCF deployment to visualise?
Limited support — you can import a VCF JSON via the Import button, which auto-populates the logical layer. Physical and storage layers need manual completion based on your hardware inventory.