NCSE-Core common mistakes
NCSE-Core · Understand the traps and how to avoid them.
Calculating usable capacity by summing raw disk capacity across nodes and applying a single replication factor, instead of accounting for the capacity reserved by the cluster's configured redundancy and the metadata/oplog overhead per node.
Candidates carry over RAID-style thinking from traditional storage and treat replication factor as a simple divisor, forgetting that RF2/RF3 in the Distributed Storage Fabric is applied at the extent/container level and that effective usable capacity differs from raw minus parity.
What to do instead
Practice the RF2 vs RF3 usable-capacity math explicitly: raw capacity minus per-node reserved space, then divided by the replication factor, and be able to explain why a 4-node RF3 cluster wastes more than a 4-node RF2 cluster. Rehearse one worked example per RF level until the arithmetic is automatic.
Confusing the roles of Prism Element and Prism Central, e.g. claiming Prism Central is required to create a storage container or that Prism Element performs cross-cluster VM management.
Both are called 'Prism', and candidates learn the UI superficially without mapping each function to its owning service. Exam items deliberately probe which plane owns cluster-local storage/VM operations versus multi-cluster management, categories, and policies.
What to do instead
Build a two-column mapping table: Prism Element owns cluster-local storage containers, local VM lifecycle, and per-cluster alerts; Prism Central owns multi-cluster inventory, categories, policies, and cross-cluster operations. Test yourself by sorting ten operations into the correct column without looking.
Treating AHV as if it were ESXi and assuming features like distributed resource scheduling, per-VM snapshots via a vCenter plugin, or vMotion-equivalent live migration work identically.
Most candidates come from VMware backgrounds and project vSphere concepts onto AHV. AHV uses Acropolis and its own scheduler, and VM lifecycle operations (clone, snapshot, live migration) are handled natively without vCenter.
What to do instead
List the five AHV operations the exam expects (create, clone, snapshot, live migrate, delete) and state for each how it is performed natively in Prism and what it depends on. Explicitly note which vSphere features have no AHV equivalent and why.
Choosing a migration method (Move, native replication, storage-only, or manual export/import) without first checking source hypervisor, network bandwidth, and whether the source can be quiesced, then defending the choice on feature richness alone.
Candidates memorize product names and pick the most capable-sounding tool. Exam scenarios are designed so the correct answer hinges on constraints such as cross-hypervisor support, downtime tolerance, and available bandwidth.
What to do instead
Adopt a decision tree: identify source hypervisor, required downtime window, and data volume, then eliminate methods that fail any constraint before comparing the survivors. Practice three scenarios with different constraints and justify each elimination in one sentence.
Assuming replication factor alone provides disaster recovery, and failing to distinguish protection domain scope (local cluster vs remote site) from snapshot and replication schedules.
RF protects against disk or node failure within a cluster, not site loss. Candidates conflate data resilience with business continuity and miss questions about protection domains, remote replication targets, and RPO/RTO implications.
What to do instead
Draw the failure hierarchy: disk failure, node failure, cluster/site loss. Map RF to the first two and protection domains plus remote replication to the third. For each level, state the RPO and RTO that the mechanism can actually deliver.
Designing a solution without running pre-sales discovery: skipping workload profiling, growth projections, IOPS/latency requirements, and rack/power constraints before recommending node models and counts.
Candidates jump to a product recommendation because sizing feels like arithmetic. Real sizing depends on workload characteristics and constraints, and exam items reward candidates who gather requirements before proposing a configuration.
What to do instead
Memorize a discovery checklist (workload types, current utilization, growth rate, performance targets, rack units, power, network ports) and apply it to every scenario question before evaluating any node configuration. If a requirement is missing, note the assumption explicitly.
Describing Flow Virtual Networking and security as if it were a perimeter firewall, missing that microsegmentation policies are applied per-VM using categories and enforced at the virtual NIC level.
Candidates map security onto familiar perimeter firewall models and overlook category-based policy application. Exam items test whether policies follow VMs across clusters and how categories drive policy scope.
What to do instead
Explain the chain out loud: VM to category to security policy to enforcement point at the vNIC. Then answer why a VM moved to another cluster keeps its policy and why a perimeter firewall cannot achieve that. Rehearse with one two-tier application example.
Confusing network segmentation constructs: mixing up which VLAN backs which network, or assuming AHV requires a separate management network for every function rather than understanding the recommended separation of management, storage, and VM traffic.
Networking questions combine Open vSwitch concepts, VLANs, and bonded interfaces, and candidates learn the terms without a mental topology. They then misassign traffic types to interfaces under exam pressure.
What to do instead
Sketch a reference topology from memory: management, backplane/storage, and VM networks, with bonds and VLAN tags labeled. Redraw it weekly until you can place each traffic type without hesitation, then answer questions against that sketch.
Answering operations questions about alerts, health checks, and upgrades by describing the UI path rather than the underlying behaviour, such as what happens to running VMs during an upgrade or how NCC checks differ from Prism alerts.
Candidates study the interface and not the operational model. Exam items ask about outcomes and sequencing, not menu locations, so UI-level recall fails when the question probes impact or prerequisites.
What to do instead
For each operational task (upgrade, NCC run, alert triage), write what changes, what is unaffected, and what must be true beforehand. Focus on impact statements such as 'VMs remain running' or 'one node at a time' rather than click paths.
Treating hybrid cloud and workload mobility as a single product feature, ignoring prerequisites like network reachability, identity, and licensing that determine whether a workload can actually move or extend.
Marketing framing makes hybrid cloud sound seamless, and candidates accept that framing. Exam scenarios test whether the candidate can identify the blocking prerequisite before recommending mobility.
What to do instead
For any mobility scenario, list the four prerequisites (network path and latency, identity/authentication, licensing, and data gravity) and check each before choosing a target. Practice identifying which single prerequisite invalidates an otherwise attractive design.