When planning an infrastructure based on VMware Cloud Foundation 9.1, the question of the minimum number of ESXi hosts often dominates early conversations. However, as cloud infrastructure architects at Q2BSTUDIO point out, this metric alone is misleading. A viable deployment is not measured solely by the number of hosts, but by the ability to maintain management services, withstand failures, and allow maintenance without disrupting business workloads. This article explores the real components that define a minimum viable VCF 9.1: storage, NSX, licensing, management, and operations, also integrating technologies such as artificial intelligence, automation, and cybersecurity.
VCF 9.1 architecture introduces a new management services layer hosted on VCF Services Runtime, consuming significant resources. The simple management model requires approximately 40 vCPU and 82 GB of RAM before considering the rest of the components. This means a four-host cluster, although the technical minimum for the installer, does not provide failover or maintenance capacity. The practical recommendation for a small production environment is to use five or six hosts when management services and workloads share the same cluster, or to separate a dedicated management cluster of four to six hosts and another for workloads. This separation, similar to the standard architecture of previous versions, offers lifecycle isolation and greater maintenance predictability.
Storage is another critical pillar. Although vSAN remains the most integrated option for greenfield deployments, VCF 9.1 supports external iSCSI, NFS, Fibre Channel, and NVMe over Fabrics through converge or import flows. However, each option implies additional team and management processes. For example, external storage requires array compatibility, fabric zones, multipathing, and recovery coordination between the VCF and storage teams. Usable vSAN capacity is reduced by protection policies, rebuild space, and operational reserves; simply adding disks is not enough. In this context, Q2BSTUDIO helps companies model these needs accurately, integrating BI/Power BI tools for capacity and cost monitoring.
NSX Manager is mandatory even if the first workload uses traditional VLANs. For a production environment, a three-node NSX Manager cluster is recommended unless reduced availability is explicitly accepted. Edge nodes, on the other hand, are optional until north-south gateway services, NAT, VPN, or firewall are needed. A typical medium Edge configuration consumes about 8 vCPU, 32 GB RAM, and 200 GB per node, adding a significant load to the management plane. Physical placement must avoid both Edges sharing the same host or switch to not nullify high availability.
Licensing in VCF 9.1 is no longer a mere final step. Broadcom requires usage reports at least every 180 days to keep licenses active. The 90-day evaluation mode is useful for testing but should not be used as a production bridge without an approved licensing plan. Companies integrating cloud services like AWS or Azure can benefit from hybrid models where VCF acts as the on-premise core, and Q2BSTUDIO provides consulting services to align these architectures with cybersecurity and regulatory compliance needs.
Observability is another fundamental piece. VCF Operations is not an optional dashboard; it must be collecting health, capacity, and configuration metrics before the first business VM is powered on. VCF Automation adds a second operational model: catalogs, projects, policies, and approvals that require ongoing maintenance. For small teams, it is better to delay Automation until dedicated staff exists, but its resource impact must be foreseen from the initial design. Log management is also often underestimated: a 30-day retention period versus one year drastically changes the storage architecture.
Finally, the human factor cannot be overlooked. A small environment needs at least two or three people with cross-cutting skills in vSphere, NSX, storage, backup, identity, and networking. Knowledge concentration in a single person is an operational risk. Medium-sized teams require three to five equivalent roles, and enterprise fleets may need more than ten dedicated professionals. The incorporation of AI agents for monitoring tasks and automated response can relieve some of the burden, but always supervised by qualified staff.
In summary, the minimum viable VCF 9.1 is not four hosts. It is a combination of compute, storage, network, licensing, external services (DNS, NTP, PKI, backup), trained personnel, and operational processes. Q2BSTUDIO offers artificial intelligence, automation, and custom software development solutions to help organizations design, implement, and operate VCF platforms that truly meet business objectives, from the first workload to cloud maturity.
