300-420 — frequently asked questions

Advertising a single default route prevents aggregation switches from learning every core change, which stabilizes the network and saves memory on critical devices.

Pushing defaults to the access layer leaves the aggregation layer with full routing tables, failing to conserve memory on the more critical aggregation switches.

Shaping controls the rate of traffic you transmit to match circuit capacity, smoothing bursts toward the provider.

Policing inbound traffic caps incoming rates, preventing non-critical flows from overwhelming buffers and starving business-critical packets.

No. EPGs are specific to Cisco ACI. SD-Access uses Virtual Networks and Scalable Group Tags (SGTs) instead.

Yes. Both SD-Access and ACI overlay VXLAN using Virtual Network Identifiers (VNID) to segment tenants or VRFs.

Proactive Insights uses ML to find serious network-wide issues, while Client 360 focuses on contextual info for a single user.

Intelligent Capture performs proactive testing by generating synthetic traffic, distinct from the alerting function of Proactive Insights.

A Point of View report consolidates environment findings and recommends platform direction during the discovery phase.

It justifies the engagement before discovery begins, so it is not a discovery-phase output.

Discovery is the empathize phase: understand the current network state and customer needs, then map Cisco innovation (C); reference architecture (B) and PPDIOO (D) are design tasks.

Establishing credibility (E) is a soft skill, but the two discovery activities are gathering current network state (A) and mapping Cisco innovation to customer needs (C).

(B) and (D) belong to the design phase; PPDIOO and reference architecture guide design discussions, while discovery focuses on understanding the customer's current state (A) and needs (C).

Apple Insights is a real DNA Assurance feature, but the question's enumerated set — Network Time Travel, proactive guided remediation, on-premise and cloud analytics — excludes it.

Overlay Virtual Transport is a fabric and transport feature used in SD-Access design, not one of the assurance analytics differentiators the question asks for.

DNA Assurance runs analytics both on-premises and in the cloud, giving organizations local data control plus cloud-scale visibility, which competitors supposedly cannot match.

Anycast RP provides RP redundancy within a single PIM-SM domain by sharing one IP address among multiple RPs. The question describes independent ASes each with a local RP, requiring inter-AS source discovery, which is MSDP's function.

MSDP peers RPs across ASes and sends Source-Active messages advertising active multicast sources. When a local RP learns of a remote source, it can relay that information to local receivers so they can join the remote stream.

The question asks to monitor device state (up/down, adjacency), not a numeric gauge. Periodic pushes the full dataset every interval regardless of need, raising shared memory and CPU load.

No. IPFIX exports sampled IP flow records, not YANG-modeled device state, so it does not provide the targeted, low-impact state data described in the scenario.

Filtering inbound on EIGRP prevents EIGRP from learning RIP routes, but doesn't control what EIGRP advertises out. To stop loops, you must stop EIGRP from advertising RIP-learned routes back into the RIP domain.

Yes, RIPv1 is classful, but the principle of avoiding route feedback remains. Filtering outbound on EIGRP ensures only native EIGRP routes are sent to RIP, keeping the domains separate.

A single domain becomes unmanageable at scale. Separate domains allow for better segmentation and localized control while SD-WAN handles inter-site transport.

Yes, logically. SD-WAN provides the overlay transport that serves as the underlay connectivity for the SD-Access fabric overlays between sites.

An exhaustive feature tour overwhelms prospects and hides the handful of features that solve their specific problems. Effective selling maps only relevant ISE capabilities to stated customer requirements.

B and E keep the pitch customer-specific: multi-vendor support reduces lock-in fears, and custom profiling proves ISE can classify the customer's own devices. C shifts the focus to the product instead of the customer.

GCM is preferred for ordinary unicast data-plane traffic. The OCG carves out multicast applications as the case where the secondary CBC mode is used, which is what this question targets.

Some community notes cite release 20.1.x and later using AES-256-GCM for multicast, but the exam follows the OCG text, which specifies CBC for multicast applications.

Scavenger-class QoS marks abnormal traffic (CS1) and drops it most aggressively only during congestion, while policing drops bursts unconditionally and can harm legitimate applications.

Queuing should sit as close to the traffic source as possible at the access edge, and distribution-to-core links are deliberately overprovisioned so queuing there is unnecessary.

Multicast routing builds distribution trees based on the source. Checking the path back to the source ensures the packet is coming from the correct direction, preventing loops.

RPF is a concept used by multicast protocols like PIM, but it relies on the underlying unicast routing table (OSPF, EIGRP, etc.) to perform the lookup.

Restoring default values is best practice as it reduces configuration complexity and potential future errors. It ensures the switch uses the vendor-defined optimal metrics for the media type.

The default STP cost for Gigabit Ethernet is 4. For FastEthernet, it is typically 19. Lower costs are preferred for root port selection.

Neither the 4-class nor the 6-class hierarchy reserves DSCP CS5, so the premium broadcast traffic would have to share Best Effort, AF31 or the EF priority queue and would lose its dedicated guarantee.

The 12-class model adds streaming, OAM, transactional and bulk data classes, each needing extra queues, policers and policy configuration on every node, which breaks the ISP's requirement for minimum resource usage and simple scaling.

The scenario only defines three marked classes — EF, AF2 and AF3 — plus unclassified traffic that lands in best effort. An 8-class model adds classes the requirements never specify, so 4 classes suffice.

IntServ uses per-flow RSVP reservation signalling, which does not scale across a full service provider core. DiffServ applies per-hop behaviour based on DSCP markings and is the standard for backbone-wide QoS.

The connected keyword only permits directly connected and summary routes to be advertised, which is the stub default; all other learned prefixes remain suppressed.

It is a route-map referenced by the stub command whose permit entries name the routes the stub router will still advertise to neighbors instead of suppressing them.

receive-only makes the router advertise nothing at all, so adjacent neighbors would receive no routes, failing the requirement outright.

A full iBGP mesh uses n(n-1)/2 sessions; with nine routers that is 9×8/2 = 36 peerings.

The question asks by how many the count was reduced, so you subtract the new total: 36 − 9 = 27.

A cut fiber causes loss of light, which the interface hardware reports as carrier loss and an immediate interface-down. OSPF reacts to that event right away, so BFD adds nothing.

Use BFD when the local interface stays up despite a remote failure, such as a failure beyond the directly connected link or a logical/indirect failure that hardware cannot signal.

No. BFD's default minimum-interval and multiplier values are too slow for sub-second detection, which is why tuning is discussed for cases where BFD is genuinely needed.

Option C describes periodic polling, which is the definition of a pull model. In a push model, the device initiates the data transfer, not the manager.

While Option A suggests limited scale, push models can actually improve scalability by reducing constant query traffic, though high-frequency pushes need careful design.

VPNv4 is a provider edge and core address family. Company A and Company B customer edge routers only exchange IPv4 unicast with their PE, so the companies themselves never run VPNv4.

No. In a single global IPv4 table both companies' routes would be visible to each other. VPNv4 with route distinguishers and route targets in the core is what prevents prefix leakage.