VCF Deployment Options with Current Homelab Resources

Current Available Resources

Compute Resources

  • 2x MS-A2 Units (arriving soon)
    • Each: 16C/32T AMD Ryzen 9, 128GB DDR5, 8.5TB storage
    • Total: 32C/64T, 256GB RAM, 17TB storage
  • 3x Intel NUC6i7KYK (existing)
    • Each: 4C/8T i7-6770HQ, 64GB DDR3, 250GB NVMe
    • Total: 12C/24T, 192GB RAM, 750GB storage
  • Combined Total: 44C/88T, 448GB RAM, 17.75TB storage

Storage Resources

  • Synology DS918+:
    • 4x 8TB drives total (2x existing + 2x new)
    • Storage Pool 1: 4TB iSCSI datastore currently serving Intel NUCs
    • Storage Pool 2: 7.3TB RAID1 (8TB raw) - Available for new workloads
    • 1GbE network connection (upgradeable to 10GbE with E10G18-T1 card)
  • vSAN Capable: MS-A2 units only (NVMe drives)
  • Storage Architecture:
    • NUCs use shared iSCSI storage (Pool 1)
    • MS-A2s have local NVMe for vSAN
    • Additional 7.3TB available for expansion (Pool 2)

Network Resources

  • Current: 1GbE infrastructure (USW-Lite-16-PoE, US-8-60W switches)
  • Future: 10GbE planned (not yet purchased)

Deployment Options Analysis

Option 1: Single MS-A2 VCF + NUCs for Workloads

Architecture:

MS-A2 #1: VCF Management Domain (partial)
├── vCenter Server (21GB RAM)
├── NSX Manager (48GB RAM)
├── SDDC Manager (24GB RAM)
└── vSAN Single Node (minimal overhead)
Total: ~93GB RAM used

NUC Cluster: Tanzu/Workload Domain
├── 3x ESXi hosts
├── Shared iSCSI storage (4TB on Synology)
├── Standalone vSphere or minimal VCF
└── Tanzu Platform for Cloud Foundry

Pros:

  • Can start immediately with one MS-A2
  • NUCs provide dedicated workload capacity
  • Separates management from workloads

Cons:

  • Cannot deploy VCF Automation (requires 96GB RAM alone)
  • Limited VCF functionality without VCFA
  • Manual lifecycle management only
  • Single point of failure for management

Feasibility: ⚠️ Partially Viable

Option 2: Dual MS-A2 VCF Management + NUCs Workload

Architecture:

MS-A2 #1 + #2: VCF Management Domain
├── Host 1: vCenter, SDDC Manager
├── Host 2: NSX Manager, VCF Automation
└── vSAN cluster (2-node with witness)

NUC Cluster: VCF Workload Domain
├── 3x ESXi hosts commissioned by VCF
├── Shared iSCSI storage (4TB on Synology)
├── Dedicated vCenter (optional)
└── Tanzu Platform deployment

Pros:

  • Full VCF functionality with all components
  • Proper vSAN configuration (2-node + witness)
  • Automated lifecycle management via VCFA
  • Clean separation of domains

Cons:

  • Must wait for both MS-A2 units
  • Requires careful resource allocation
  • Network bandwidth constraints on 1GbE

Feasibility: ✅ Fully Viable

Option 3: Mixed Architecture - All 5 Hosts for VCF

Architecture:

VCF Management Domain (5 hosts)
├── MS-A2 #1: vCenter, VCF Automation (primary)
├── MS-A2 #2: NSX Manager, SDDC Manager
├── NUC #1-3: Additional capacity, vSAN contributors
└── Mixed vSAN cluster (challenging)

Pros:

  • Maximum resource pool (448GB RAM total)
  • Distributed workload across all hosts
  • No separate domains needed

Cons:

  • VCF doesn’t officially support mixed CPU vendors
  • vSAN performance mismatch (NVMe vs SATA)
  • USB NIC reliability concerns on NUCs
  • Complex troubleshooting

Feasibility: ❌ Not Recommended

Option 4: Standalone vSphere + Tanzu (No VCF)

Architecture:

vSphere Environment
├── MS-A2 #1: vCenter Server, Management
├── MS-A2 #2: NSX Manager, Tanzu Supervisor
├── NUC Cluster: Tanzu Workload Cluster
└── Manual NSX-T deployment

Pros:

  • More flexible resource allocation
  • No VCF licensing requirements
  • Can deploy components selectively
  • Start immediately

Cons:

  • No unified management (SDDC Manager)
  • Manual lifecycle management
  • More complex Day 2 operations
  • No VCF automation benefits

Feasibility: ✅ Fully Viable

Option 5: Nested VCF Lab on MS-A2s

Architecture:

Physical MS-A2 Hosts
├── MS-A2 #1: Nested ESXi VMs for VCF
├── MS-A2 #2: Additional nested hosts
└── NUCs: Workload capacity (physical)

Pros:

  • Learn full VCF without hardware limitations
  • Can simulate larger environments
  • Flexibility to test configurations

Cons:

  • Performance overhead from nesting
  • Not suitable for production workloads
  • Complex networking setup
  • High memory consumption

Feasibility: ✅ Viable for Learning

Option 6: Phased Deployment Approach

Phase 1: Foundation (1 MS-A2)

MS-A2 #1: Basic vSphere + NSX-T
├── vCenter Server
├── NSX Manager (standalone)
└── Prepare for VCF migration

Phase 2: VCF Introduction (2 MS-A2s)

Import to VCF Management Domain
├── Deploy SDDC Manager
├── Import existing vCenter/NSX
└── Add VCF Automation

Phase 3: Full Stack (2 MS-A2s + 3 NUCs)

Complete VCF Deployment
├── Management Domain (MS-A2s)
├── Workload Domain (NUCs)
└── Tanzu Platform for CF

Pros:

  • Start working immediately
  • Learn progressively
  • Minimize downtime during transitions
  • Spread costs over time

Cons:

  • Requires migration/rebuild at Phase 2
  • Some rework involved
  • Temporary limitations

Feasibility: ✅ Highly Recommended

Resource Allocation Comparison

VCF Component Requirements vs Available Resources

Deployment Option Total vCPU Needed Available vCPU Total RAM Needed Available RAM Viable?
Single MS-A2 VCF 86 cores 16 cores 312GB 128GB ❌ No
Single MS-A2 (No VCFA) 62 cores 16 cores 216GB 128GB ⚠️ Partial
Dual MS-A2 VCF 86 cores 32 cores 312GB 256GB ⚠️ Tight
Dual MS-A2 + NUCs 86 cores 44 cores 312GB 448GB ✅ Yes
Standalone vSphere Flexible 44 cores Flexible 448GB ✅ Yes

Decision Matrix

Criteria Option 1 Option 2 Option 3 Option 4 Option 5 Option 6
Can Start Now ✅ ❌ ❌ ✅ ✅ ✅
Full VCF Features ❌ ✅ ⚠️ ❌ ⚠️ ✅*
Resource Efficiency ⚠️ ✅ ❌ ✅ ❌ ✅
Complexity Medium Low High Medium High Medium
Production Ready ❌ ✅ ❌ ✅ ❌ ✅
Learning Value ⚠️ ✅ ⚠️ ✅ ✅ ✅
Cost Low Low Low Low Low Low

*Eventually, after Phase 2

Recommendations

Immediate Action (This Week)

Go with Option 6 - Phased Deployment

  1. Deploy vCenter on first MS-A2
  2. Install ESXi on all Intel NUCs
  3. Create basic vSphere cluster
  4. Begin learning NSX-T manually

Short Term (When 2nd MS-A2 Arrives)

Transition to Option 2

  1. Deploy VCF across both MS-A2 units
  2. Import existing configuration
  3. Commission NUCs as workload domain
  4. Full VCF experience

Long Term (With 10GbE)

  • Optimize network performance
  • Enable vMotion across domains
  • Better vSAN performance
  • Production-ready environment

Special Considerations

William Lam’s Single-Host VCF

  • Requires JSON deployment method
  • Skip VCF Automation component
  • Use for learning only
  • Not recommended for workloads

Network Bandwidth Impact

  • 1GbE limits vSAN performance
  • vMotion will be slower
  • NSX overlay adds overhead
  • Plan for 10GbE upgrade
  • iSCSI storage traffic competes with other network traffic on 1GbE

Storage Architecture Considerations

Current NUC Storage Setup:

  • NUCs boot from local 250GB NVMe
  • VMs stored on 4TB iSCSI datastore (Synology DS918+)
  • Shared storage enables vMotion between NUCs
  • 1GbE network limits storage performance

Benefits of Mixed Storage:

  • MS-A2s use local NVMe for performance-critical VCF components
  • NUCs leverage existing iSCSI investment
  • Separation of storage domains (vSAN vs iSCSI)
  • Future 10GbE upgrade will significantly improve iSCSI performance

Storage Allocation Strategy:

  • Pool 1 (4TB): Continue serving Intel NUCs workloads
  • Pool 2 (7.3TB): Options include:
    • Backup/archive storage for VCF components
    • ISO/template repository
    • Additional VM storage for MS-A2 hosts
    • Tanzu persistent volumes
    • vSphere Content Library

Licensing Requirements

  • VCF requires specific licenses
  • vSphere licenses may differ
  • Consider vExpert/VMUG advantages
  • Plan for license allocation

This analysis provides a comprehensive view of all viable options with your current resources, helping you make an informed decision based on your immediate needs and long-term goals.

This project is for educational and home lab purposes.