VMware VCP-DCV: Data Center Virtualization Study Guide
The VMware VCP-DCV (Data Center Virtualization) certification validates the ability to install, configure, optimize, and troubleshoot vSphere-based data center solutions built on ESXi and vCenter Server. The exam runs 130 minutes, requires a scaled passing score of 600, and is aimed at administrators and engineers who design and operate production vSphere environments. Expect scenario-driven questions spanning architecture, VMware product selection, design decisions, configuration steps, performance tuning, troubleshooting, and day-to-day operations.
Domain 1: Architecture and Technologies
- ESXi is a Type-1 (bare-metal) hypervisor: its purpose-built VMkernel installs directly on physical hardware and schedules CPU, memory, storage, and devices for guest VMs with no underlying general-purpose host OS, unlike a Type-2 hypervisor that runs as an application on a host OS.
- The Virtual Machine Monitor (VMM) is the per-VM component that virtualizes the guest's CPU and memory and works with the VMkernel; hardware-assisted virtualization (Intel VT-x / AMD-V) offloads CPU virtualization to the processor for efficiency.
- A vSphere Standard Switch (vSS) is configured and managed per host, while a vSphere Distributed Switch (vDS) is created and managed centrally in vCenter Server, providing consistent network configuration and advanced features across all hosts in a cluster.
- Aligning a VM's virtual NUMA (vNUMA) topology with the physical NUMA architecture minimizes remote memory access; a VM whose memory spans multiple NUMA nodes incurs higher remote-access latency.
- VMFS (Virtual Machine File System) is the clustered file system for block storage that lets multiple ESXi hosts read/write the same datastore concurrently to support vMotion, HA, and DRS.
- Thick Provision Eager Zeroed allocates and zeroes the entire disk at creation time and is required for Fault Tolerance and certain clustered (MSCS-style) applications; Thick Lazy Zeroed and Thin defer zeroing until first write.
- vSphere Fault Tolerance (FT) in vSphere 7 and 8 supports VMs with up to 8 vCPUs by maintaining a secondary shadow VM on another host in lockstep with the primary.
- DirectPath I/O (PCI passthrough) and VMDirectPath I/O give a VM direct, dedicated access to a physical PCIe device such as a GPU or NIC, bypassing the hypervisor's device emulation but generally disabling vMotion for that VM.
- Enhanced vMotion Compatibility (EVC) masks newer CPU features across mixed processor generations from the same vendor so VMs can vMotion freely within a cluster; EVC does not work across Intel/AMD boundaries.
- VM Encryption protects VMDKs and VM files and requires a Key Provider in vCenter, either an external KMS (Standard Key Provider) or the built-in vSphere Native Key Provider.
- In vSphere with Tanzu, each Kubernetes pod can run inside a lightweight vSphere Pod that provides VM-level hardware isolation while integrating with the Supervisor Cluster.
- vCenter Server is deployed only as the Linux-based vCenter Server Appliance (VCSA); the standalone external Platform Services Controller (PSC) has been deprecated in favor of an embedded PSC.
- NVMe-oF (NVMe over Fabrics) delivers lower latency and higher IOPS than SCSI-based Fibre Channel by using a streamlined command set and parallelism designed for flash storage.
- When DRS is set to manual or disabled, or a host enters partial maintenance, DRS stops issuing migration recommendations and initiates no new vMotion operations, but already-running VMs continue to run unaffected.
Domain 2: VMware Products and Solutions
- VMware vSAN is the hyperconverged storage solution that pools local NVMe/SSD/HDD across ESXi hosts into a single distributed datastore; the Failures to Tolerate (FTT) policy defines how many host/disk failures objects can survive.
- vSAN's Express Storage Architecture (ESA) replaces the legacy disk-group model (cache + capacity tiers) with a single flat pool of NVMe devices managed per host.
- VMware Aria Operations (formerly vRealize Operations) provides intelligent operations management: performance monitoring, capacity planning, and optimization for vSphere environments.
- VMware Aria Operations for Logs (formerly vRealize Log Insight) is the log aggregation and analytics appliance; ESXi hosts forward syslog to it over the standard syslog protocol.
- VMware Aria Operations for Networks (formerly vRealize Network Insight) provides network visibility, flow analysis, and micro-segmentation planning.
- VMware Aria Automation (formerly vRealize Automation) delivers self-service provisioning, infrastructure-as-code, and cloud automation.
- VMware vSphere Replication provides host-based, hypervisor-level asynchronous replication of individual VMs without dependence on array-based replication.
- VMware Site Recovery Manager (SRM) automates disaster recovery orchestration, runbooks, and failover/failback, typically working with vSphere Replication or array-based replication.
- vSphere Lifecycle Manager (vLCM) provides unified, desired-state management of ESXi base images, vendor add-ons, drivers, and firmware for whole-cluster patching and updates.
- vSphere with Tanzu enables Workload Management: enabling a Supervisor Cluster turns ESXi hosts into a Kubernetes control plane for running and managing containerized applications.
- VMware NSX provides network virtualization with distributed logical routing, overlay networking, and a distributed firewall for micro-segmentation independent of physical topology.
- VMware HCX (Hybrid Cloud Extension) enables large-scale workload mobility and migration between on-prem and cloud sites with bulk and live (vMotion-based) migration capabilities.
- vCenter Server is the centralized management platform for ESXi hosts, VMs, and vSphere services; it enables clustering features such as HA, DRS, and vMotion.
- Legacy migration into vSphere can use VMware vCenter Converter Standalone for physical-to-virtual conversions, though HCX or third-party tools are preferred for modern large-scale migrations.
Domain 3: Planning and Designing
- A vSAN stretched cluster requires a dedicated witness host (or witness appliance) deployed at a third site to act as a tiebreaker and prevent split-brain when the two data sites cannot communicate.
- With the standard FTT=1 RAID-1 mirroring policy, vSAN consumes raw capacity at roughly a 2:1 ratio (each object stores two copies plus a witness component).
- Use Storage Policy-Based Management (SPBM) to define per-tier storage policies (for example FTT, stripe width, and space reservation) and assign them to VMs or disks for differentiated service levels.
- Place high-IOPS database workloads on NVMe-backed VMFS 6 datastores or an all-flash vSAN configuration, and separate contending workloads onto different datastores to avoid I/O contention.
- HA admission control can reserve failover capacity using a static number of hosts, a cluster resource percentage, or dedicated failover hosts; the percentage policy is recommended for clusters with dissimilar host sizes.
- Set critical VMs to the Highest HA restart priority and enable VM Component Protection (VMCP) so HA responds to All Paths Down (APD) and Permanent Device Loss (PDL) datastore-accessibility failures.
- vMotion requires a VMkernel adapter tagged for vMotion traffic on every participating host, plus shared storage (for traditional vMotion) accessible to all hosts in the cluster.
- For multi-vCenter management, deploy two or more VCSAs in Enhanced Linked Mode to get a single pane of glass and shared roles, tags, and licenses across instances.
- Use VM-VM anti-affinity (Separate Virtual Machines) rules to keep redundant VMs (such as clustered web servers) on different hosts, and affinity rules to keep tightly coupled VMs together.
- For network design, separate traffic types (management, vMotion, vSAN, VM) onto different VLANs and VMkernel adapters, and use explicit failover order across uplinks to isolate traffic onto distinct active/standby NICs.
- Use LACP or static EtherChannel on a distributed switch only when the physical switch is configured to match; otherwise rely on load-based teaming such as Route based on physical NIC load.
- For containerized workloads, plan a supported networking stack such as NSX or the NSX Advanced Load Balancer to provide load balancing and pod networking for the Supervisor Cluster.
- Size latency-sensitive or large VMs to fit within a single physical NUMA node when possible (for example matching vCPU count to a socket's core count) to avoid remote memory access.
- Combine vSphere HA for automatic VM restart, FT for continuous availability of the most critical small VMs, and DRS for load balancing and non-disruptive maintenance migrations.
Domain 4: Installing and Configuring
- VCSA deployment is a two-stage process: Stage 1 deploys the OVF appliance, and Stage 2 performs configuration including SSO domain (for example vsphere.local), NTP, and the administrator@<domain> password.
- If a server's storage/RAID controller driver is not in the default ESXi installation ISO, build a custom ESXi image (Image Builder) that includes the vendor driver VIB, or the installer will not detect local disks.
- ESXi requires GPT (GUID Partition Table) for boot and system disks; MBR is legacy and limited.
- vLCM uses a desired-state image specification consisting of an ESXi base image version, a vendor add-on, firmware, and additional drivers that all hosts in the cluster must match; remediation enforces compliance.
- Remediating non-compliant hosts with vLCM places each host into maintenance mode, migrates its VMs via DRS, applies the reference image, and reboots, processed sequentially across the cluster.
- A VM's virtual hardware (compatibility) version determines the available virtual devices and maximum resource limits; pick the version compatible with the lowest ESXi version in the cluster.
- Software iSCSI uses a software iSCSI adapter (a vmhba) with target servers added via dynamic (SendTargets) or static discovery; bind VMkernel ports for multipathing.
- For VM-to-VM security on a port group, set Promiscuous Mode, MAC Address Changes, and Forged Transmits to Reject to prevent sniffing and spoofing.
- Content libraries: create a Published (local) library with publishing enabled at the source, then create a Subscribed library elsewhere pointing to the published library's subscription URL to share templates, ISOs, and OVFs.
- Auto Deploy provisions stateless ESXi hosts via PXE boot combined with host profiles to apply a consistent reference configuration at scale.
- VM Encryption configuration requires a Key Provider (external KMS Standard Key Provider or the vSphere Native Key Provider) registered in vCenter before VMs can be encrypted.
- The default vDS uplink-port and teaming policy is Route based on originating virtual port; choose the teaming policy that matches the physical network and redundancy requirements.
- vSphere HA needs at least one shared datastore reachable by all cluster hosts for datastore heartbeating, used to distinguish a host isolation/partition from an actual host failure.
- Upgrade order is vCenter Server first, then ESXi hosts, because a newer ESXi version is not supported when managed by an older vCenter.
Domain 5: Performance-tuning and Optimization
- High CPU Ready time means a VM's vCPUs are waiting for physical CPU scheduling due to contention; remediate by reducing the VM's vCPU count or rebalancing workloads across hosts.
- Use the VMXNET3 paravirtualized virtual NIC (not E1000/E1000E) for best network performance; it supports multi-queue, RSS, large offloads, and lower CPU overhead.
- Use the Paravirtual SCSI (PVSCSI) controller for high-I/O VMs to reduce CPU overhead and increase throughput compared to the default LSI Logic controller.
- A CPU reservation guarantees a minimum amount of CPU (MHz) a VM always receives; CPU shares only determine relative priority during contention (a VM with 2x shares gets roughly twice the CPU time of a Normal-share VM when CPU is contended).
- A memory limit caps the physical RAM a VM can use regardless of configured size; if the limit (for example 4 GB) is below the guest's demand, the host forces ballooning or swapping for the remainder, hurting performance.
- Set a VM's Latency Sensitivity to High to have the VMkernel exclusively reserve CPU resources and disable ballooning/TPS for that VM, reducing jitter for latency-critical workloads (requires full CPU reservation).
- Storage I/O Control (SIOC) enforces datastore-wide fairness by throttling lower-priority VMs (per disk shares) when device latency exceeds the congestion threshold.
- Network I/O Control (NIOC) on a vDS uses bandwidth allocation, shares, and reservations across network resource pools to prioritize traffic types such as vMotion, vSAN, and VM traffic.
- Storage DRS (SDRS) automatically balances VM disk placement and migrations across a datastore cluster based on used-space and I/O-latency metrics, and handles intelligent initial placement.
- The DRS migration threshold (aggressiveness slider) controls how readily DRS recommends or performs vMotion migrations to balance cluster load.
- High KAVG (kernel average latency) indicates the ESXi VMkernel storage stack is the bottleneck, typically from device queue saturation or HBA queue-depth limits; high DAVG points to the array/device, GAVG is the guest-observed total.
- Sustained mem.swapused.average greater than zero indicates the host is swapping VM memory to disk, a sign of memory overcommitment and a serious performance problem.
- On a hybrid vSAN, write latency caused by destaging that cannot keep up with incoming writes overflows the SSD cache tier; move to all-flash or reduce write pressure to resolve it.
- Content-Based Read Cache (CBRC), exposed to administrators as View Storage Accelerator, caches common read blocks in host RAM to accelerate boot-storm read workloads such as VDI; this is distinct from vSphere Flash Read Cache (vFRC), which caches to local host SSD.
Domain 6: Troubleshooting and Repairing
- When a host shows as isolated or partitioned in HA, its management network lost connectivity to other cluster members and the isolation address; verify physical management network and the configured isolation address.
- A host stops sending HA heartbeats when its management network fails; confirm network connectivity between the host and vCenter and between hosts before assuming a host failure.
- If a VM did not restart after a host failure, check that its HA restart priority was not set to Disabled and that admission control left enough failover capacity.
- For a VM stuck on a specific VLAN, verify the VLAN is allowed (trunked) on the physical switch port connected to the ESXi host uplink and that the port group VLAN ID is correct.
- Consolidating or deleting large snapshots causes heavy I/O and can briefly stun the VM as delta (redo-log) disks are merged into the base VMDK; quiescing the guest filesystem adds further delay.
- An orphaned snapshot chain (deltas remain after a failed delete) is fixed via right-click VM > Snapshots > Consolidate to commit the deltas into the base disk.
- If FT cannot be enabled, check for blockers such as an existing snapshot chain, attached USB devices, physical-mode RDMs, or unsupported virtual hardware.
- A vMotion that fails to start may be blocked by a connected CD-ROM backed by a local ISO on the source host, or by device passthrough; disconnect host-local resources first.
- All Paths Down (APD) means every storage path to a datastore is unavailable but the device may return; Permanent Device Loss (PDL) means the array signaled the device is gone for good and triggers VMCP action.
- After a driver-caused NIC or HBA issue, check the VMware Compatibility Guide (HCL) for a supported driver/firmware version, update the VIB, and open a vendor support case if it persists.
- Reset access to a host directly through the DCUI (Direct Console User Interface) for tasks like restarting management agents, resetting networking, or re-enabling lockdown recovery.
- Key vCenter logs live in /var/log/vmware/vpxd/vpxd.log; for a sick VCSA, check appliance storage usage and PostgreSQL (vmware-vpostgres) status and restart the vmon-managed service stack if needed.
- When VMs cannot power on due to insufficient resources, review HA admission control settings and either add cluster capacity or relax the failover level/percentage.
- If DRS is not migrating VMs off an overloaded host, check for VM-Host affinity rules, per-VM DRS automation overrides, or DRS being set to manual that block automatic vMotion.
Domain 7: Administrative and Operational Tasks
- To create a reusable template, run Sysprep (or equivalent guest customization prep) inside the guest, shut down the VM, then right-click and Convert to Template; deploying a VM from it applies a customization specification.
- A Full Clone creates a complete, independent copy of all VM files, while a Linked Clone shares the parent's base disk and stores only deltas, saving space at the cost of dependency on the parent.
- Implement least-privilege access by creating a custom role with the needed VM privileges, assigning it to a group, and granting that permission at the appropriate inventory object (such as a VM folder).
- A permission defined directly on a child object overrides a permission propagated from a parent object for that specific object.
- Use proportional shares for resource pools (for example an 8000:2000, or 4:1, ratio for production vs. development) so each pool gets CPU/memory in proportion during contention.
- Make the VMware Certificate Authority (VMCA) a subordinate CA of the enterprise CA so VMCA-issued machine and solution certificates chain up to the corporate root of trust.
- Configure VCSA file-based backup in the VAMI (appliance management interface on port 5480) by scheduling backups to an SFTP/FTPS/NFS/SMB target.
- Cross-vCenter vMotion moves a VM between vCenter instances; with Enhanced Linked Mode it can change compute, storage, and network simultaneously, and Advanced Cross vCenter vMotion works even between unlinked vCenters.
- A combined 'Change both compute resource and storage' vMotion moves a running VM to a new host and a new datastore at the same time, requiring both source and destination datastores be accessible from the running host.
- Snapshot consolidation commits the changes in all delta disks into the base VMDK and then removes the delta files, reclaiming the space they occupied.
- Per-VM EVC mode lets an individual VM carry its CPU feature baseline as it migrates, enabling consistent migration across clusters or to the cloud independent of cluster EVC.
- Changing a VM's vSAN storage policy triggers a non-disruptive background resynchronization that rebuilds object components to satisfy the new policy.
- When using array auto-tiering under a datastore cluster, disable the SDRS I/O metric (I/O load balancing) so SDRS space balancing does not conflict with the array's tiering decisions.
- Before raising a VM's hardware compatibility level, take a snapshot first since the upgrade is not reversible without one; the VM must be powered off to apply the change.
VMware VCP-DCV exam tips
- VCP-DCV questions are heavily scenario-based: read for the specific constraint (latency-sensitive, regulated/encrypted, mixed CPU generations, third site) because that detail usually points directly to the one correct feature.
- Memorize the current product naming: vRealize products are now VMware Aria (Operations, Operations for Logs, Operations for Networks, Automation); the exam may use either name.
- Know the diagnostic metrics cold: CPU Ready (scheduling contention), KAVG/DAVG/GAVG (storage stack vs. device vs. guest), ballooning and mem.swapused (memory pressure), and what each tells you to fix.
- Be precise about reservations vs. limits vs. shares: a reservation guarantees a minimum, a limit caps the maximum, and shares only matter during active contention.
- Watch sequencing answers: vCenter is upgraded before ESXi, a Key Provider is configured before VM encryption, and vLCM remediation always uses maintenance mode plus DRS migration before each host reboot.
Study guide FAQ
How long is the VCP-DCV exam and what score do I need to pass?
The exam runs 130 minutes and uses a scaled scoring model where 600 is the passing score (scores range roughly 100-500 for fail and 600 for pass on VMware's scale). It is multiple choice and multiple response, with no hands-on lab in the core exam.
Which vSphere version should I study for?
Study the current vSphere 8 platform while remaining comfortable with vSphere 7 behavior, since many concepts (vLCM desired-state images, FT supporting up to 8 vCPUs, embedded PSC, vSAN ESA) carry across both. Avoid memorizing deprecated items like the external standalone PSC except to recognize that they are deprecated.
How much of the exam is troubleshooting versus design and configuration?
Troubleshooting and Repairing, Installing and Configuring, Architecture and Technologies, and Performance-tuning are the largest areas, so weight your prep toward hands-on operational skills: reading logs and metrics, fixing HA/network/storage issues, and choosing correct configuration steps rather than pure theory.
Do I need real lab experience to pass, or is reading enough?
Hands-on experience is strongly recommended because the questions test practical judgment, such as interpreting KAVG/CPU Ready, ordering vLCM remediation, and resolving snapshot or vMotion blockers. A home lab or nested ESXi environment plus the official VMware documentation will reinforce the facts far better than reading alone.