I85PREP
⌘ K
Dell VxRail Operate study materials
Dell VxRail Operate study resources

Dell VxRail Operate: VxRail Architecture and Deployment Fundamentals

Dell VxRail Operate · Work through your exam, one topic at a time.

All guide sections

VxRail Architecture and Deployment Fundamentals

VxRail is Dell's jointly engineered hyperconverged platform: PowerEdge-based nodes running VMware ESXi and vSAN, with VxRail Manager as the appliance's control point. This domain tests whether you know what VxRail owns versus what vSphere owns, and the topology rules that decide whether a proposed cluster is even valid.

Nodes and the software stack

  • Models are grouped by workload: E-series balanced, P-series performance (NVMe-heavy, GPU options), V-series for VDI, S-series storage-dense, D-series dense, G-series GPU. Each model pairs with a validated release bundle.
  • ESXi boots from a redundant BOSS card or mirrored M.2 modules, never from a vSAN device.
  • vSAN consumes dedicated cache and capacity devices; the cache tier absorbs writes, capacity holds data.
  • iDRAC provides per-node out-of-band management; VxRail Manager does not replace it.

Cluster topology rules

  • A standard cluster starts at three nodes; two-node clusters require a witness, and stretched clusters need at least two nodes per site plus a witness at a third location.
  • 64 nodes per cluster is the vSAN ceiling VxRail designs against.
  • RAID-5 erasure coding needs four hosts, RAID-6 needs six; below those counts the only FTT>0 option is mirroring.
  • VxRail Manager is a VM on the cluster, so it is unavailable if the cluster is down - this drives recovery design.

Worked example

A customer asks for a stretched cluster with three hosts in the primary site and one in the secondary, claiming it "still survives a site loss". That design is invalid: the secondary site holds a single host, so losing it removes any component copy placed there and policy compliance collapses. Valid options are 2+2+witness at minimum, or 3+3+witness, with inter-site round-trip latency inside vSAN's 5 ms RTT limit. The witness holds metadata and quorum only - it protects against split-brain, not against data loss.

Traps that cost marks

  • Treating VxRail Manager as the vCenter Server; they are separate VMs with separate roles.
  • Assuming the witness stores VM data or can be a repurposed VxRail node.
  • Quoting RAID-5 for a three-node cluster.
  • Forgetting that growing a two-node cluster changes the witness requirement.
  • Assuming a model's release bundle matches another model's.