An architect is designing the network architecture for a multi-site VMware Cloud Foundation (VCF) deployment using NSX Federation. The environment spans two geographically dispersed data centers (Site A and Site B) to support disaster recovery, workload mobility, and consistent security policies across sites. During the design workshop, the architect determined that the solution must leverage NSX Federation for stretched networking to support overlay transport, routing, and high availability.
Which five design decisions are required when implementing NSX Federation in a multi-site VCF environment? (Choose five.)
A. Configure a stretched Global Tier-0 Gateway for North/South connectivity of spanning segments with BG
B. Limit round-trip latency between sites to 150 ms or less to enable optimal workload mobility and policy synchronization.
C. Ensure end-to-end MTU of at least 1,700 bytes on the underlay network to support Geneve encapsulation for federated overlays.
D. Require NSX Edge nodes at both sites for all federated constructs, even when using Distributed Transit Gateways for VPCs.
E. Configure separate local Tier-1 gateways for site-specific segments, in addition to a spanning Tier-1 for cross-site communication.
F. Deploy a Global Manager in one site to centrally configure and synchronize spanning Tier-0 and Tier-1 gateways.
G. Use DHCP IP pools for host and Edge Tunnel Endpoints (TEPs).
H. Ensure end-to-end MTU of at least 1,500 bytes on the underlay network to support Geneve encapsulation for federated overlays.
Explanation:
NSX Federation uses a Global Manager to centrally define and synchronize global networking and security objects across the Local Managers at participating sites, making F fundamental to the architecture. Broadcom confirms that configuration is performed through the active Global Manager and synchronized to the applicable locations. (Broadcom TechDocs)
A is required for stretched North/South routing: a Global Tier-0 can span locations, with site-local service routers and BGP connectivity to the physical network. Federation also uses RTEPs on Edge Nodes for cross-site Geneve transport.
B reflects the important 150-ms cross-location Edge-to-Edge RTT limit for stretched networking. Newer NSX releases permit higher latency between some manager components, but the inter-site Edge data path remains subject to the tighter latency consideration.
C is correct because NSX overlay networking requires additional MTU for Geneve encapsulation. Broadcom specifies 1600 bytes as the minimum and 1700 bytes as the recommended value, making 1700 the appropriate design target.
E separates site-local routing from stretched workloads: local Tier-1 gateways serve site-specific segments, while a Global/spanning Tier-1 provides cross-site connectivity. Broadcom confirms that local Tier-1 gateways can connect to a Global Tier-0 in Federation.
D is too absolute, particularly with distributed VPC connectivity models. G is incorrect because Edge TEP addressing is normally based on NSX IP pools/static allocation rather than requiring DHCP. H provides insufficient MTU for Geneve overlays.
Study Guide References/Topics: NSX Federation Architecture; Global and Local Managers; Stretched Tier-0/Tier-1 Gateways; RTEPs; Multi-Site BGP Routing; Geneve MTU; Local versus Global Networking Objects; Multi-Site Availability.