Academy/VCAP — VCF Networking (3V0-25.25)
VCF

VCAP — VCF Networking (3V0-25.25)

VCF 9.0vcap-advanced

Master NSX federation, advanced routing with BGP/OSPF, micro-segmentation at scale, ALB integration, and multi-site networking for large enterprise deployments. Deep focus on operational resiliency and automation.

3V0-25.25
VCAP
60
Questions
135m
Duration
300 (scaled score; Broadcom publishes no percentage or ceiling)
Pass Score
30
Objectives

Exam Blueprint Weights

Section titles, groupings and weights below are VCDX Academy study groupings, NOT the official Broadcom blueprint structure. Broadcom publishes no section weights. Always cross-check the official exam guide. Official exam guide ↗
Section 1 — Architecture
~15%
Section 2 — Routing Design
~25%High Weight
Section 3 — Security Services
~20%
Section 4 — Load Balancing and Network Services
~15%
Section 5 — Troubleshooting
~25%High Weight

Version Evolution

VCAP Networking covers advanced NSX design and operations. Evolution mirrors NSX itself: NSX-V edge-centric → NSX-T with T0/T1 model → NSX 4.x with federation and advanced security. VCF 9.0 aligns NSX versioning with VCF (NSX 9.0.x). Exam focuses on multi-site federation, edge design, and integration with Avi ALB for load balancing.

Learning Outcomes

  • Understand Advanced VCF 9.0 Networking (NSX & Advanced Routing) concepts and architecture
  • Design multicast on NSX with PIM sparse mode and IGMP on the Tier-0 external interfaces (rendezvous point on the physical fabric), and recognise that multiprotocol BGP (SAFI 2) is a fabric-side topic that did not surface in the VCF 9.0.1 NSX routing documentation.
  • Prevent inter-VRF route leakage on a multi-VRF Tier-0 with per-VRF prefix lists and route maps on the BGP neighbours and redistribution, and verify advertisements with 'get bgp neighbor summary' / 'advertised-routes' on the Edge.
  • IPSec and L2 VPN design: choose policy-based vs route-based IPSec, IKE version/profile and compliance suite, place the service on the right gateway (Active-Standby), and position L2 VPN against Federation stretched segments for migration and DR.

Advanced NSX Design & Multi-Site Federation#

NSX Form Factors & Sizing

NSX deployment sizing depends on workload scale, geographic distribution, and performance SLAs:

NSX Manager Form Factors
Small, Medium, Large and Extra Large size the NSX Manager VM itself (vCPU, memory, disk); they are not deployment tiers with bundled Edge counts, and Edge nodes are sized separately (below). The highly available model uses a three-node NSX Manager cluster, of which only one node can fail at a time (VCF 9.0.1 docs p.2773); a single NSX Manager node appears only in the VCF simple deployment model, with slower recovery and limited scale (VCF-NSX-LM-REQD-CFG-004, p.561). Do not use Small for a Global Manager (p.2673).
Scale Limits
Hosts, segments and rules per deployment come from the VMware Configuration Maximums tool for the NSX release; check it rather than memorising figures.

Edge VM Sizing (identical in NSX 4.2 and NSX 9.0; disk is 200 GB for every size)
: Small 2 vCPU / 4 GB - proof-of-concept ONLY, T1 L7 rules not realized. Medium 4 vCPU / 8 GB - L2-L4, under 2 Gbps. Large 8 vCPU / 32 GB - L2-L4 at 2-10 Gbps, plus L7 LB and TLS inspection. Extra Large 16 vCPU / 64 GB - high-throughput L7 LB, VPN, URL filtering, IDS/IPS. Broadcom does not designate one size 'production-recommended' - size by throughput and feature band. DPDK is a property of the host CPU (AESNI + 1 GB huge pages), not of the Edge form factor.

Manager Node Cluster
3 nodes for HA; quorum is a majority (2 of 3). The cluster self-heals after one node failure and stays available with two nodes; if two nodes fail, write operations and clustering API/CLI calls are blocked until a majority is restored (VCF 9.0.1, Troubleshoot NSX Manager Cluster). Place the three nodes on separate hosts (anti-affinity) on the management domain vSAN.

NSX Federation: Multi-Site Architecture

Federation allows multiple NSX domains (sites) to operate as unified network fabric:

Global Manager
Single point of management for the Federation. An NSX Federation environment contains one active and one standby Global Manager cluster (3 nodes each) plus one or more Local Manager clusters; the standby takes over by manual or automatic activation. Global Manager supports Policy mode only. Size the GM nodes from the NSX Manager form-factor table (VCF 9.0.1, Configuring the Global Manager and Local Managers; NSX Global Manager Model).
Local Managers
One per site; subordinate to Global Manager. Independent operation if Global Manager unavailable (graceful degradation).
Stretched Segments
Logical switch synchronized across sites. VM can vMotion between sites while maintaining IP/MAC. The Global Manager is management plane only: it pushes configuration to the Local Managers (VCF 9.0.1 docs p.2293); for DHCP, Federation supports DHCP Relay on segments and gateways, and a DHCPv4 server on gateways with DHCP static bindings configured on segments (Features Supported in NSX Federation table).
Location-Aware Groups
DFW rules scoped per location. Example: block-malware group can differ per site; Global Manager compiles per-location rules.
Gateway Failover
Tier-0 gateways at each site peer upstream; stretched failover follows the Tier-0 HA mode: active-active with primary and secondary locations, active-active with all primary locations, or active-standby with primary and secondary locations (VCF 9.0.1 docs p.2314).
┌──────────────────────────────────────────────────────────────┐
│                      Global Manager                          │
│         (one active + one standby GM cluster)               │
└───────────────────────┬──────────────────────────────────────┘
          ┌─────────────┴─────────────┐
          │                           │
    ┌─────▼─────┐             ┌───────▼────┐
    │  Site-A   │             │  Site-B    │
    │  Local Mgr│             │  Local Mgr │
    │  3x ESXi  │             │  3x ESXi   │
    │ Tier-0 GW │             │ Tier-0 GW  │
    └─────┬─────┘             └────┬───────┘
          │ BGP Peer               │ BGP Peer
          │ (to core)              │ (to core)
    ┌─────┴─────────────────────────┴──────┐
    │        IP WAN / L3 Underlay          │
    └───────────────────────────────────────┘
  • Stretched Segments: Segment-A exists on both sites
  • DFW Rules: Location-aware rules apply per site
  • Service VLAN: Unique per site; GSLB load balances across

Host Switch — VDS Only, N-VDS Deprecated

Two migrations are routinely confused on exam questions; keep them apart:

NSX-V to NSX-T (V2T, historic)
NSX for vSphere (edge-centric, VXLAN) to NSX-T (Tier-0/Tier-1, Geneve) using the Migration Coordinator. VCF 9.0 estates do not contain NSX-V; the topic survives only as background.
N-VDS (NSX Virtual Distributed Switch)
An NSX-T host switch type, not an NSX-V construct. Deprecated in NSX 4.0: ESX hosts can only be prepared as transport nodes on a vSphere Distributed Switch (VDS 7.0 or later, vCenter 7.0 or later). N-VDS remains only as the switch inside an NSX Edge VM transport node (VCF 9.0.1, Host Switches; Managing NSX on a vSphere Distributed Switch).
N-VDS to VDS host-switch conversion
Applied to clusters that were prepared on N-VDS under NSX-T 3.x before an upgrade to 4.x: the host switch is converted to a VDS and segments are realized as NSX distributed virtual port groups in vCenter; VMs stay on their segments. It is a host-switch change, not a VM-by-VM vMotion between two NSX domains.
VCF context
SDDC Manager/VCF Installer prepares every VCF 5.x and 9.0 workload domain on VDS; vSphere Lifecycle Manager images require VDS. A VCDX-level design states VDS explicitly and treats any surviving N-VDS cluster as a pre-upgrade remediation item.

Transport Node Profiles & NIOC

Transport Node Profiles standardize hypervisor configuration at scale:

Transport Node Profile
Template defining VMKs (Virtual Machine Kernel adapters), VLAN transport zones, overlay transport zones. Applied to cluster or host level.
Network I/O Control (NIOC)
QoS for VMK traffic. Example: vSAN traffic guaranteed 20% bandwidth, vMotion 10%, management 5%, VM data best-effort. Enforces fairness across competing workloads.
MTU Sizing
Geneve overlay minimum MTU is 1600; 1700 is recommended and is the NSX default Global TEP MTU (headroom for future Geneve options). VCF design rule VCF-NET-REQD-OVL-001: overlay segment MTU = TEP MTU minus 200. Where the fabric supports jumbo frames, 9000 is recommended. Validate end-to-end path supports MTU.
Uplink Profiles
Define active/standby topology, failover strategy (active-active requires LAG or ECMP), LLDP/CDP for underlay discovery.
# Where Transport Node Profiles are created (NSX Manager UI or Policy API) — there is no
# interactive "create transport-node-profile" CLI on NSX Manager.

UI : System > Fabric > Hosts > Transport Node Profile > Add Transport Node Profile
       Host Switch  = VDS (select vCenter + VDS)      Mode = Standard
       Transport Zones = <overlay TZ>, <VLAN TZ>      Uplink Profile = <profile, Global TEP MTU 1700>
       TEP IP assignment = IP Pool / DHCP              Uplink-to-VDS-uplink mapping
     System > Fabric > Hosts > Clusters > Configure NSX  → apply the profile to the vSphere cluster

API: PUT /policy/api/v1/infra/host-transport-node-profiles/<profile-id>      (PolicyHostTransportNodeProfile)
     PUT /policy/api/v1/infra/sites/default/enforcement-points/default/transport-node-collections/<id>
         (TransportNodeCollection = profile → cluster binding)

Verify (NSX Manager/Edge NSX CLI, admin):
     get transport-nodes status
     get transport-node <node-uuid> status

In VCF the profile is created and applied by SDDC Manager (5.x) or VCF Installer/Fleet Management (9.0) when the domain or cluster is created; operators edit it only through the VCF workflows.

Key Takeaways

  • For exam: NSX Manager form factors (Small/Medium/Large/Extra Large) size the Manager VM in a three-node cluster; take scale limits from the VMware Configuration Maximums tool. Edge VM sizing: Small 2vCPU/4GB (PoC only), Medium 4/8 (<2 Gbps), Large 8/32 (2-10 Gbps + L7 LB/TLS), XL 16/64 (L7 LB, VPN, IDS/IPS). All 200 GB disk.
  • For exam: Federation architecture: one active + one standby Global Manager cluster (3 nodes each, Policy mode only) + a Local Manager cluster per site; NSX Manager cluster quorum is a majority (2 of 3). Stretched segments enable VM vMotion while maintaining IP/MAC. Location-aware DFW rules differ per site.
  • For exam: Transport Node Profile template for VMKs, VLAN/overlay zones. NIOC enforces QoS fairness (vSAN 20%, vMotion 10%, management 5%). MTU: 1600 minimum for Geneve overlay, 1700 recommended (NSX default TEP MTU); 9000 where the fabric supports jumbo frames. Fabric MTU should exceed workload MTU by at least 100 bytes, 200 preferred.

Advanced Routing: BGP, OSPF, and Route Redistribution#

BGP Deep Dive — Route Maps, Prefix Lists, Community Strings

Border Gateway Protocol is critical for multi-site, high-availability networking:

BGP Basics
Autonomous System Number (ASN) per site. Tier-0 router acts as BGP speaker; advertises pod CIDR, service CIDR, stretched segment CIDR to upstream DC fabric.
Route Maps
Policy-based routing logic. Example: Tag pod routes with community 65000:1000 (production), pod routes with 65000:1001 (test). Upstream routers use communities to apply different QoS per workload.
Prefix Lists
Define allowed IP ranges for incoming/outgoing routes. Prevent accidental advertisement of 0.0.0.0/0 or private ranges.
AS Path Prepend
Artificially inflate AS path to deprioritize certain routes. Example: prepend 65000 3 times to make backup site route less preferred; upstream chooses primary site by shortest AS path.
Local Preference (LocalPref)
BGP attribute 0-4294967295. Higher value = preferred. Set high LocalPref on primary site, lower on secondary. Overrides AS path length.
Route Redistribution
Inject Tier-0/Tier-1 connected routes (segments, uplinks, loopbacks, service IPs, NAT, LB VIPs, IPSec local endpoints) and static routes into BGP, filtered by a redistribution route map. NSX does not redistribute between BGP and OSPF in either direction (VCF 9.0.1, Configure an NSX OSPF).
# Where BGP policy objects live in NSX (UI or Policy API). The route map is then referenced
# at the BGP neighbour (in/out filter) or in Route Re-distribution — not typed on a CLI.

UI : Networking > Tier-0 Gateways > <tier0-a> > Edit > Routing
       IP Prefix Lists  : prod-prefixes  = 10.100.0.0/16 Permit, 10.200.0.0/16 Permit, any Deny
       Community Lists  : prod-comm      = 65000:1000 ; test-comm = 65000:1001
       Route Maps       : prod-tag       = seq 10 match IP Prefix prod-prefixes → set Community 65000:1000, Permit
                                           seq 20 match IP Prefix test-prefixes → set Community 65000:1001, Permit
     Networking > Tier-0 Gateways > <tier0-a> > BGP > BGP Neighbors > 10.0.0.1
       Remote AS 65001 · Address family IPv4 · Out Filter = prod-tag · Max Prefixes (Maximum Routes) 10000
     Networking > Tier-0 Gateways > <tier0-a> > Route Re-distribution > Set → choose sources, attach route map

API: PATCH /policy/api/v1/infra/tier-0s/tier0-a/prefix-lists/prod-prefixes          (PrefixList)
     PATCH /policy/api/v1/infra/tier-0s/tier0-a/community-lists/prod-comm            (CommunityList)
     PATCH /policy/api/v1/infra/tier-0s/tier0-a/route-maps/prod-tag                  (Tier0RouteMap)
     PATCH /policy/api/v1/infra/tier-0s/tier0-a/locale-services/default/bgp/neighbors/upstream-1
           (BgpNeighborConfig: neighbor_address, remote_as_num, route_filtering[].out_route_filters=[prod-tag])

Verify (Edge node NSX CLI, admin):
     get logical-routers                               # find the Tier-0 SR vrf id
     vrf <id>
     get bgp neighbor summary                          # state, prefixes received/sent
     get bgp neighbor 10.0.0.1 advertised-routes       # KB-noted: use this if the Out counter looks wrong

Route-map set actions available in NSX: AS-path prepend, MED, Weight (0-65535, higher preferred, local to the Edge), Community (regular aa:nn, large aa:bb:cc, NO_EXPORT / NO_ADVERTISE / NO_EXPORT_SUBCONFED), Local Preference (highest preferred). Match criteria: IP prefix list or community list (MATCH ANY/ALL/EXACT/REGEXP) (VCF 9.0.1, Create an NSX Route Map).

OSPF on the Tier-0 — What NSX Actually Supports

OSPF is an alternative northbound protocol between the Tier-0 and the physical fabric where the fabric team runs OSPF rather than BGP. Its NSX feature envelope is narrow (VCF 9.0.1, Configure an NSX OSPF, p.1877):

Version and placement
OSPFv2 only (no OSPFv3), configured on a Tier-0 gateway (default VRF only, not on Tier-0 VRF or Tier-1). Active-active or active-standby Tier-0; up to two Tier-0 uplinks per Edge node run OSPF.
Areas
A single area per Tier-0 (so a Tier-0 is never an ABR). Area types: backbone (0), normal, or NSSA. Stub and totally-stubby areas are not offered. MD5 or plain-text authentication per area.
Redistribution
The same sources as BGP (uplinks, downlinks, loopbacks, Tier-1 downlinks, service routes) are redistributed as Type 5 external LSAs (normal area) or Type 7 (NSSA), always type-2 metric (E2/N2); route summarisation of E2/N2 routes is supported. Redistribution route maps match prefix lists only, with no set actions. No redistribution between BGP and OSPF.
Coexistence with BGP
OSPF and BGP may run together only for BGP multi-hop, where the BGP peer address is learned through OSPF.
Resilience
BFD-backed sessions; graceful restart helper mode only; default-originate supported; ECMP up to 8 paths. Interfaces are broadcast or numbered point-to-point (/31).
Not supported
NAT with OSPF, Tier-0 HA VIP with OSPF, NSX Federation with OSPF.
Router ID
Highest interface IP unless a loopback exists; adding a loopback after OSPF has chosen a RID does not change it until OSPF is disabled/enabled or the Edge is redeployed.
Protocol fundamentals that still appear on the exam (RFC 2328, not NSX-specific)
Cost = reference bandwidth / interface bandwidth; with the conventional 100 Mbps reference, a 1 GbE and a 10 GbE link both round to cost 1, so set an explicit interface cost or a higher reference bandwidth on the physical routers when you need link preference. OSPF floods LSAs (an LSP is the IS-IS term); each LSA is refreshed every 30 minutes (LSRefreshTime) and ages out at 60 minutes (MaxAge). Convergence is driven by hello/dead timers (10 s/40 s on broadcast interfaces by default) or by BFD, not by the LSA refresh interval.
Design verdict for VCF
VCF Validated Solutions and the VCF 9.0 design guidance use eBGP between Tier-0 and the fabric. Choose OSPF only where the fabric mandates it, accept single-area/no-VRF/no-Federation limits, and document them as design constraints.

Route Redistribution at Scale

What NSX redistributes, and where policy is applied:

Sources
Tier-0: external interface subnets, loopbacks, static routes, connected segments, service interface subnets, NAT IPs, LB VIP/SNAT IPs, IPSec local endpoints, DNS forwarder IPs, EVPN TEP. Tier-1 (advertised to the Tier-0, then redistributed northbound): connected segments, static routes, NAT IPs, LB VIP/SNAT IPs, IPSec local endpoints, DNS forwarder IPs.
Static ← redistribute
A static default route or ISP-learned static on the Tier-0 can be redistributed into BGP (or OSPF as E2/N2). Prefer a floating static (higher administrative distance) only as a backup to a dynamically learned default.
Filtering
Attach a redistribution route map to permit only the intended prefixes. Example: deny the management subnet 10.255.255.0/24 while permitting the workload aggregates, so an untrusted eBGP peer never learns it; apply a second route map as the neighbour out-filter for defence in depth.
Path preference across sites
Use BGP attributes, not metric translation: Local Preference (highest wins, inbound policy on your own routers), AS-path prepend on the backup site's outbound route map, MED to an external peer. Weight (0-65535) is Edge-local and never propagates. There is no OSPF-cost-to-BGP mapping because NSX never redistributes between the two.

ECMP & Multi-Path Load Balancing

Maximizing throughput via parallel paths:

8-Way ECMP
Tier-0 can load-balance across up to 8 equal-cost paths. Requires 8 edge nodes with active-active uplinks.
Hash Algorithm
Northbound Tier-0 ECMP spreads flows across up to 8 equal-cost paths. On ESX hosts, the ECMP 5-tuple feature (protocol, source and destination address, source and destination port) is disabled by default and is enabled by setting lb_ecmp to true through the NSX Policy API (VCF 9.0.1 docs p.1985).
Asymmetric Paths
Return traffic may take different path than outbound (valid in NSX). Ensure firewall rules symmetric; no hairpinning required.
Link Redundancy
8 edge nodes at same site all contribute to ECMP pool. Loss of 1 edge = 12.5% throughput reduction (1/8 paths). Rebalance load automatically.

IPv6 & Dual-Stack Design

VCF 9.0 supports IPv6 at Tier-0/Tier-1 level:

Dual-Stack Segments
Logical switch with both IPv4 and IPv6 CIDR. DHCPv6 server can auto-assign /64 to pods.
Router Advertisement (RA)
Tier-1 router issues RA packets; stateless autoconfig for VM networks. Stateful DHCPv6 alternative.
BGP IPv6
Separate address family (AFI 2, SAFI 1) in BGP config. Tier-0 advertises both IPv4 and IPv6 routes independently.
NAT64
NSX NAT64 is stateful: binding and session information is kept on the Edge service router, and only an IPv6-only client can initiate to an IPv4-only server (unicast TCP/UDP). An IPv4 pool is mapped to a /96 IPv6 prefix on the Tier-0/Tier-1 (VCF 9.0.1, Add an NSX NAT64).

Key Takeaways

  • For exam: BGP route maps + communities tag production (65000:1000) vs. test routes (65000:1001) for upstream QoS. AS path prepend deprioritizes backup routes. LocalPref 0-4294967295 overrides AS path length.
  • For exam: NSX OSPF = OSPFv2 only, Tier-0 default VRF only, one area per Tier-0 (backbone/normal/NSSA, no stub), redistributed routes are E2/N2 (Type 5/7), no BGP↔OSPF redistribution, no Federation, ECMP up to 8 paths. OSPF cost = reference bandwidth / link bandwidth (1 GbE and 10 GbE both = 1 at the 100 Mbps reference); LSAs refresh every 30 minutes; convergence comes from hello/dead timers or BFD.
  • For exam: ECMP up to 8 equal-cost paths. On ESX hosts 5-tuple (L4) ECMP hashing is off by default (lb_ecmp=true enables it). Loss of 1 edge = 12.5% throughput reduction. Asymmetric paths valid; ensure firewall rule symmetry.

Advanced BGP Engineering#

BGP Session Types and Scalability

BGP session design determines convergence time, route scalability, and failover behavior. NSX edges support three primary architectures: iBGP full-mesh (N squared sessions, complex), route-reflector (star topology, simpler), and confederation (hierarchical for multi-region).

Flow-Specific Routes and BGP Routing Policies

Advanced scenarios require route filtering and re-distribution:

Key Takeaways

  • IPv4 Multicast: NSX multicast routing is PIM sparse mode on Tier-0 external interfaces with IGMP (hello/hold timers, IGMP join as a debugging tool — VCF 9.0.1 p.1849); the rendezvous point lives in the physical fabric. Multiprotocol BGP for multicast (SAFI 2) did not surface in the VCF 9.0.1 NSX documentation — treat it as the physical-fabric side of the design, not an NSX feature.
  • BGP Route Policies to Prevent VRF Leak:In multi-VRF deployments, accidentally leak routes from high-security VRF to low-security VRF = security breach. Apply export policies on Tier-0: per-VRF route maps denying routes from forbidden VRFs. Monitor route advertisements: get bgp neighbor summary (Edge NSX CLI, inside vrf <tier0-sr-id>)

Multi-VRF on Tier-0#

VRF-Lite and VRF Attachment Strategy

VRF-Lite enables multiple independent routing tables on a single Tier-0. Each VRF has separate BGP session, route table, and interface namespace. Use case: multi-tenant deployments (each tenant = VRF), PCI zone isolation (payment processing VRF separate from general VRF), OT/IT isolation (operational tech on VRF-OT, IT on VRF-IT).

Key Takeaways

  • For exam: VRF-Lite = separate routing tables + BGP sessions on single Tier-0. Each VRF independent route table and interface namespace. Use for multi-tenant, PCI isolation, OT/IT segregation.
  • For exam: Route leak vulnerability: high-security VRF → low-security VRF = breach. Mitigate with export route maps per VRF, deny forbidden VRF routes. Monitor: get bgp neighbor summary (Edge NSX CLI, inside vrf <tier0-sr-id>).
  • For exam: VRF-Lite more efficient than multi-site federation for single-site isolation. Federation for stretched segments and GSLB across geographic locations.

Advanced NSX Federation#

Global Manager Cluster Architecture: RAFT-based Consensus

NSX Federation enables multi-site deployments with global object management. Global Manager (GM) cluster (3 nodes in RAFT consensus) centralizes policy, stretched segments, and global groups. Local Managers (one per site) register with GM, enforce global policies locally.

Key Takeaways

  • For exam: Global Manager cluster: 3 nodes RAFT consensus. Local Managers per site register with GM and enforce global policies locally. Graceful degradation if GM unavailable.
  • For exam: Stretched segments synchronized across sites. VM vMotion maintains IP/MAC. The active Global Manager synchronises configuration to the relevant Local Managers (VCF 9.0.1 docs p.2293); in Federation, DHCP runs as DHCP Relay on segments and gateways or as a DHCPv4 server on gateways with static bindings on segments (p.2292). Location-aware DFW rules compiled per-site by GM.
  • For exam: Stretched Tier-0 failover follows the Edge HA mode (active-active with primary/secondary or all-primary locations, or active-standby). Tier-0 gateways peer upstream per site. GSLB load balances service VIPs across sites.

Dataplane Performance and DPDK Tuning#

Edge VM CPU Pinning and Hugepages

NSX edge performance depends on dataplane (DP) core allocation and pinning. Edges run control plane (CP) on one core (management, BGP, SSH), dataplane on N cores (packet forwarding).

Key Takeaways

  • For exam: NSX edge architecture: 1 CP core (management, BGP, SSH) + N DP cores (packet forwarding). CPU pinning mandatory for predictable performance. Hugepages reduce TLB misses.
  • For exam: DPDK (Data Plane Development Kit) enables high-throughput packet processing. CPU isolation avoids context switches. Monitor: NSX Manager → System → Performance → Edge.
  • For exam: DP core saturation = packet drops. Scale horizontally: add more edge nodes. Verify uplink capacity matches DP throughput (no bottleneck upstream).

Enterprise Integrations with Third-Party Infrastructure#

Cisco ACI Border Leaf Peering

NSX Tier-0 can establish eBGP sessions with Cisco ACI border leafs, enabling multi-vendor fabric. NSX advertises its routes (workload subnets), ACI advertises fabric routes.

Key Takeaways

  • For exam: Cisco ACI integration: eBGP session NSX Tier-0 ↔ ACI border leaf. NSX advertises workload subnets; ACI advertises fabric routes. ASN differs (NSX vs. ACI), enables eBGP dynamic peering.
  • For exam: Multi-vendor fabric: validate BGP convergence, route leak prevention (route maps, prefix lists). Test failover: disable one border leaf, verify alternative path.
  • For exam: Integration complexity trade-off: ACI + NSX = advanced segmentation but operational overhead. Single vendor (NSX-only) simpler to manage but less flexibility if ACI already deployed.

IPv6 in NSX: Dual-Stack and Transition Mechanisms#

Dual-Stack Segment Support and IPv6 Routing

NSX supports dual-stack segments: both IPv4 and IPv6 subnets on same logical switch. Tier-1 gateways use static routes and advertise their connected IPv6 segments to the Tier-0; dynamic routing (BGP IPv6 address family) runs on the Tier-0 only. NSX supports OSPFv2 only, so IPv6 northbound routing is BGP or static (VCF 9.0.1, Configure an NSX OSPF: 'Only OSPFv2 is supported').

Key Takeaways

  • For exam: Dual-stack segment = IPv4 + IPv6 CIDR on same logical switch. DHCPv6 auto-assign /64 to pods. Router Advertisement (RA) enables stateless autoconfig vs. stateful DHCPv6.
  • For exam: BGP IPv6: separate AFI 2 / SAFI 1. Tier-0 advertises IPv4 and IPv6 routes independently. No OSPFv3 in NSX (OSPFv2 only): IPv6 northbound routing is BGP or static on the Tier-0.
  • For exam: NAT64 is stateful (binding/session state on the Edge SR), IPv6-only client initiates to IPv4-only server over unicast TCP/UDP; IPv4 pool mapped to a /96 prefix; configured on a Tier-0/Tier-1 with a service router.

VPN Deep Dive: IPSec, SSL, and Layer-2 Stretch#

IPSec VPN — Policy-Based vs Route-Based, IKE, Profiles

NSX Edge terminates two VPN types, IPSec VPN (site-to-site) and Layer 2 VPN. Both are configured under Networking > VPN in NSX Manager or through the Policy API, on a Tier-0, Tier-0 VRF or Tier-1 gateway running in Active-Standby HA mode (VCF 9.0.1, Virtual Private Network (VPN), pp.2108-2134).

Transport
Tunnel mode with ESP (IP protocol 50); IKE on UDP 500, UDP 4500 when NAT is detected. Pre-shared key or certificate authentication (RSA with SHA-256/384/512).
Policy-based IPSec
Local and peer subnets define what is protected; the policy must be mirrored at the remote site. Static by nature; no tunnel redundancy; DNAT is not supported on a gateway carrying a policy-based session.
Route-based IPSec
A virtual tunnel interface (VTI) per tunnel; traffic is steered by static routes or BGP over the VTI. Tunnel redundancy (primary/failover tunnels) is supported on a Tier-0 with BGP only; OSPF over IPSec is not supported, and dynamic routing on the VTI is not supported on Tier-1 gateways.
IKE versions
IKE profile selects IKEv1, IKEv2 (default), or IKE-Flex (initiate IKEv2, accept IKEv1 from the peer; no fall-back initiation). IKEv1 is still a selectable option, not removed. System defaults: nsx-default-l3vpn-ike-profile and nsx-default-l2vpn-ike-profile = IKEv2, AES-CBC-128, SHA2-256, DH group 14. Compliance suites (CNSA, FIPS, Foundation, PRIME, Suite-B-GCM-128/256) pre-set IKE/IPSec parameters that cannot be edited.
Objects to build
VPN Service (IPSec) on the gateway → Local Endpoint (tunnel source IP: uplink or loopback, with the certificate for cert-auth) → IKE profile, IPSec (tunnel) profile and DPD profile (defaults exist) → IPSec Session (policy- or route-based) referencing all of the above. Download Config on the session gives the remote-side parameters.
Scale (Table 862)
Medium Edge: 128 IPSec tunnels per policy-based VPN session, 128 sessions per VPN service, 2048 tunnels per VPN service. Large Edge: 128 tunnels per policy-based session (soft limit), 256 sessions and 4096 tunnels per VPN service (16 tunnels per session). Small Edge: not for production VPN.
High availability
The IPSec state is synchronised to the standby Edge; on failover the tunnels continue without renegotiation for both policy- and route-based sessions.

Remote-Access (SSL) VPN — Not an NSX Feature

The NSX VPN chapter in VCF 9.0.1 documents IPSec VPN and L2 VPN only. Client-to-site SSL VPN (SSL VPN-Plus) was an NSX for vSphere Edge feature and has no equivalent on NSX Tier-0/Tier-1 gateways. Remote-user access in a VCF design is delivered by a third-party or cloud access gateway, a partner NFV appliance on an NSX segment, or Workspace ONE/VDI, and the design states that choice rather than assuming an NSX SSL VPN endpoint.

Layer 2 VPN — Stretching a Segment Over IPSec

L2 VPN extends L2 networks (VNI- or VLAN-backed) across sites in one broadcast domain, so VMs keep their IP addresses when moved. It is supported and documented in NSX 9.0 — it is not deprecated — and remains the tool for migrations and onboarding where NSX Federation is not in place.

Mechanics
An L2 VPN server and an L2 VPN client; the GRE tunnel is carried inside a route-based IPSec tunnel. One L2 VPN service (client or server) per Tier-0 or Tier-1 gateway; one GRE tunnel per session; up to 4094 segments per session; no tunnel redundancy. VLAN↔VNI, VLAN↔VLAN and VNI↔VNI extension; VLAN trunking over VDS 7.0+. Path MTU discovery is on by default (PMTU cache 10 minutes). NSX-to-NSX only, with managed, standalone/unmanaged (VLAN only) and autonomous Edge clients.
When to use which
NSX Federation stretched segments for a standing multi-site design under one Global Manager; L2 VPN for migration waves, cloud-provider onboarding and DR designs where the sites are not federated or the remote side is an autonomous Edge. Flag the broadcast-domain stretch (storms, no tunnel redundancy) as a design risk in either case.

Key Takeaways

  • For exam: NSX VPN = IPSec (policy-based or route-based) and L2 VPN only; no SSL/remote-access VPN on NSX 9.0. VPN services sit on a Tier-0, Tier-0 VRF or Tier-1 in Active-Standby mode; ESP tunnel mode, IKE UDP 500/4500.
  • For exam: IKE profile versions IKEv1 / IKEv2 (default) / IKE-Flex; defaults nsx-default-l3vpn-ike-profile and -l2vpn- = IKEv2, AES-CBC-128, SHA2-256, DH 14; compliance suites fix the parameters. Route-based tunnel redundancy needs BGP on a Tier-0; OSPF over IPSec is not supported.
  • For exam: L2 VPN rides a route-based IPSec tunnel (GRE inside), one service per gateway, 4094 segments per session, no tunnel redundancy, NSX-to-NSX only. Federation stretched segments are the standing multi-site answer; L2 VPN is the migration/onboarding answer.

Network Security Engineering at Scale#

DFW Rule Management: 10k+ Rules Operational Excellence

Mature NSX deployments grow to tens of thousands of DFW rules. Managing that scale requires discipline, automation, and performance tuning.

Lab Exercise: Design and Deploy Multi-Tier Segmentation

Key Takeaways

  • For exam: DFW rule hierarchy: category → policy → rules. Categories are processed in order (Ethernet, Emergency, Infrastructure, Environment, Application; vDefend DFW docs, Distributed Firewall Categories); rules are evaluated top-to-bottom and the first match wins. The default rule allows traffic (VCF 9.0.1 docs p.1524: the default DFW rules allow all Layer 2 and Layer 3 traffic) unless you set an explicit default deny.
  • For exam: 10k+ rules optimization: break into sections per tier (management, app, db, edge). Use groups instead of individual IPs (membership updates push to all rules). Monitor rule hit rate.
  • For exam: Performance tuning: the DFW is already stateful (it tracks flows; VCF 9.0.1 docs p.3621), so nothing needs enabling; prefer L3 + L4 rules over L7 inspection (lower CPU). Test failover: verify rule sync across edges.

Exam Mapping: 3V0-25.25 — Advanced VCF 9.0 Networking (NSX & Advanced Routing)

  • See Advanced VCF 9.0 Networking (NSX & Advanced Routing) exam blueprint for detailed objectives

Labs in This Section

📝 Quiz — VCAP Advanced Networking

0/50 correct

Architecture

Q1
NSX overlay uses Geneve encapsulation which requires the physical underlay to support what minimum MTU to avoid fragmentation?
  • 1280
  • 1600
  • 1500
  • 9000 required
NSX overlay uses Geneve encapsulation which adds ~50 bytes of header overhead to each frame. The physical underlay must support minimum 1600 MTU to avoid fragmentation. 1500 is standard (too small). 9000 is jumbo (recommended but not required). 1280 is IPv6 minimum.
Q2
BFD in NSX overlay is used to:
  • Replace BGP
  • Perform IDS inspection
  • Encrypt tunnels
  • Quickly detect TEP-to-TEP tunnel failures
BFD (Bidirectional Forwarding Detection) quickly detects TEP-to-TEP tunnel failures at sub-second intervals, enabling fast failover. It doesn't encrypt tunnels, replace BGP, or perform IDS inspection.
Q3
The NSX Management Plane in a cluster provides which functions?
  • Edge data plane forwarding
  • Only firewall enforcement
  • Only BGP routing
  • Manager and Policy APIs with integrated controller services
The NSX Management Plane cluster provides Manager and Policy APIs with integrated controller services. Controller functionality was merged into the Manager starting in NSX-T 3.x. It doesn't only enforce firewalls, handle BGP routing, or do Edge forwarding.
Q4
DPU-based acceleration offloads:
  • vCenter transactions
  • BGP policy calculation
  • NSX data-plane functions (e.g., overlay/DFW) to a SmartNIC
  • Storage I/O to vSAN
DPU-based acceleration offloads NSX data-plane functions (overlay encap/decap, DFW enforcement) to SmartNICs, freeing host CPU. It doesn't offload vCenter transactions, vSAN I/O, or BGP calculations.
Q5
Which transport zone type carries Geneve overlay traffic?
  • Global transport zone
  • Edge-only zone
  • VLAN transport zone
  • Overlay transport zone
Overlay transport zones carry Geneve-encapsulated overlay traffic. VLAN transport zones carry VLAN-tagged traffic. Global and Edge-only zones are not standard NSX transport zone types.

Routing Design

Security Services

Load Balancing and Network Services

Troubleshooting

🃏 Flashcards — VCAP Advanced Networking

52 cards
Card 1 of 52
Central Control Plane (CCP)
The NSX control plane function inside the NSX Manager cluster responsible for computing runtime state (segment tables, TEP lists) from policy and distributing it to transport nodes. It is distinct from the management plane that stores intent. CCP health is critical to fast convergence.
Was this page useful?
Type to search. ↑ ↓ to move, Enter to open, Esc to close.