Skip to content

VMware Cloud on AWS (VMC on AWS)

Scope

VMware Cloud on AWS (VMC on AWS): host types (i3.metal, i3en.metal, i4i.metal), SDDC sizing, stretched clusters, linked VPC networking, HCX migration, Elastic vSAN, Elastic DRS, NSX overlay design, and AWS service integration.

VMware-managed SDDC running on dedicated AWS bare-metal infrastructure. Provides full VMware stack (vSphere, vSAN, NSX, vCenter) on AWS with native AWS service integration.

Checklist

  • [Critical] SDDC size: single-host (dev/test only, no SLA) vs. multi-host (production, minimum 2 hosts)?
  • [Critical] Host type selection: i3.metal (36 vCPU, 512GB RAM, 15.2TB NVMe), i3en.metal (larger storage), or i4i.metal (latest gen)?
  • [Critical] Stretched cluster (multi-AZ) for HA, or single-AZ SDDC? Stretched clusters require minimum 6 hosts (3 per AZ + 2 witness).
  • [Critical] Networking connectivity: linked VPC for AWS service access, Transit Connect for multi-VPC, Direct Connect for on-prem?
  • [Critical] Migration strategy: HCX bulk migration, live vMotion, or replication-assisted vMotion from on-prem?
  • [Critical] Pricing model: on-demand (per-host hourly) or 1yr/3yr subscription (significant discount)?
  • [Recommended] vSAN storage policy: dedup+compression enabled, FTT level (failures to tolerate), RAID-1 vs. RAID-5/6?
  • [Recommended] Elastic vSAN (disaggregated storage) for workloads needing more storage than compute?
  • [Recommended] Elastic DRS configuration: auto-scale hosts based on utilization thresholds, min/max host count?
  • [Recommended] NSX overlay network design: segments, firewall rules, NAT, VPN between SDDC and on-prem?
  • [Recommended] AWS service integration plan: which AWS services (RDS, S3, Lambda, ELB) will VMware VMs consume via linked VPC?
  • [Recommended] Add-ons needed: SRM for DR, Tanzu for Kubernetes, VMware Aria for operations?
  • [Optional] Content library configuration for VM templates and ISOs across SDDCs?
  • [Optional] Multi-SDDC or SDDC group topology for large deployments with shared networking?
  • [Optional] Hybrid Linked Mode to manage on-prem vCenter and VMC vCenter from single pane?

Why This Matters

VMC on AWS eliminates physical infrastructure management while preserving VMware operational familiarity. Misconfigurations are expensive: a single i3.metal host costs ~$8-9/hour on-demand. Choosing single-host dev SDDCs for production workloads means no vSAN redundancy and no SLA. Not using Direct Connect for high-throughput on-prem connectivity leads to unpredictable latency over public internet. Failing to right-size host count wastes significant budget since you pay per-host regardless of VM utilization.

Common Decisions (ADR Triggers)

Decision When to Create ADR
Single-AZ vs. stretched cluster Always — defines HA posture and cost (stretched doubles minimum hosts)
Host type selection Always — determines compute/storage ratio and per-host cost
On-demand vs. subscription pricing Always — 1yr saves ~30%, 3yr saves ~50%, but commits capacity
Linked VPC vs. Transit Connect When VMs need AWS services — linked VPC is simpler, Transit Connect scales to multi-VPC
HCX migration approach When migrating from on-prem — bulk vs. vMotion affects downtime windows
Elastic DRS enablement When workloads have variable demand — auto-scaling hosts affects cost predictability
SRM vs. native AWS DR When DR is required — SRM preserves VMware DR runbooks, AWS-native requires redesign
Elastic vSAN adoption When storage/compute ratio is imbalanced — disaggregated storage adds flexibility but complexity

Reference Architectures

  • DC Extension: on-prem vSphere + VMC SDDC connected via Direct Connect + HCX, stretched L2 networking for seamless migration
  • DR Landing Zone: on-prem primary site, VMC as DR target using SRM, pilot-light or warm-standby hosts scaled up during failover
  • Migration Factory: VMC as landing zone for large-scale DC migration, HCX bulk migration waves, post-migration optimization with Elastic DRS
  • Hybrid Application: VMware VMs for stateful workloads + AWS-native services (RDS, S3, Lambda) via linked VPC for new microservices
  • Dev/Test in Cloud: single-host SDDCs for non-production, multi-host production SDDC, separate SDDCs per environment

Key Constraints

  • VMware manages ESXi hosts — no custom ESXi images or direct host access
  • Minimum 2 hosts for production SLA; single-host is dev/test only with no redundancy
  • SDDC is in a dedicated AWS account managed by VMware; linked VPC connects to your account
  • vCenter, NSX Manager, HCX Manager consume resources on management hosts
  • Host scaling is not instant — adding hosts takes 15-20 minutes
  • All VMware licensing (vSphere, vSAN, NSX, vCenter) included in per-host price

See Also

  • providers/vmware/infrastructure.md -- VMware vSphere and VCF infrastructure
  • providers/vmware/networking.md -- NSX networking design patterns
  • providers/vmware/avs-azure.md -- Azure VMware Solution as alternative cloud VMware
  • providers/vmware/gcve-gcp.md -- Google Cloud VMware Engine as alternative cloud VMware