Comprehensive and Detailed 150 to 250 words of Explanation From [SR Linux EVPN and Data Center Interconnect/Course Guide/topics]:

A decoupled gateway-based DCI model separates the data center border-leaf function from the WAN PE function. This separation is the key design point. The border leaf remains part of the data center EVPN/VXLAN environment, while the WAN PE participates in WAN VPN transport and policy enforcement. Because the roles are split across two devices, the handoff between the border leaf and WAN PE provides a clean administrative and operational boundary. That boundary is useful for security policy, QoS marking, traffic classification, and troubleshooting ownership. The WAN does not need direct reachability to every leaf and route reflector as in a gateway-less model. The WAN PE also does not peer directly with the data center route reflector in a decoupled model; route exchange occurs through the border-leaf/WAN-PE handoff. VXLAN tunnels between leaf routers across different data centers are characteristic of gateway-less extension, not decoupled gateway operation. Therefore, the statement about clear demarcation between the data center border leaf and WAN PE is the accurate description. Reference: decoupled gateway-based DCI, security/QoS demarcation, WAN PE separation.

================

Comprehensive and Detailed 150 to 250 words of Explanation From [SR Linux EVPN and Data Center Interconnect/Course Guide/topics]:

In an all-active Layer 2 EVPN multi-homing design, the host is normally dual-attached through a LAG to multiple leaf routers that share the same Ethernet Segment Identifier. Leaf1 and Leaf2 both participate in the Ethernet Segment and may receive traffic from the host. For BUM traffic sourced by the host, the host-side hashing can send frames toward either attached leaf. For BUM traffic sent from the EVPN overlay toward the multi-homed segment, DF election controls which PE forwards that replicated traffic toward the local Ethernet Segment to prevent duplicate delivery. The false statement is option B. A remote leaf such as Leaf3 does not simply enable ECMP on the MAC-VRF to load-balance traffic between Leaf1 and Leaf2. EVPN all-active forwarding uses Ethernet Segment discovery, Ethernet A-D routes, aliasing, and split-horizon procedures to determine valid next-hops and prevent loops. ECMP alone is an underlay or routing-table behavior; it is not the MAC-VRF mechanism that authorizes multi-homed L2 forwarding across an Ethernet Segment. Reference: all-active L2 EVPN multi-homing, Ethernet Segment association, DF election, aliasing.

================

Comprehensive and Detailed 150 to 250 words of Explanation From [SR Linux EVPN and Data Center Interconnect/Course Guide/topics]:

An anycast-gateway IRB allows multiple PEs to present the same default-gateway IP address to hosts in the same subnet. This is the mechanism that enables distributed gateway behavior in EVPN fabrics. The same anycast gateway IP may be configured on equivalent IRB subinterfaces across remote PEs participating in the same IP-VRF, allowing hosts to use the nearest leaf as their default gateway without changing their gateway address. SR Linux also associates gateway MAC information with the IRB, including virtual gateway MAC behavior used inside the MAC-VRF forwarding table. Option C is false because the anycast gateway IP is not treated as a normal unique host route that appears in the IP-VRF route table alongside the subnet prefix. The subnet route is installed for the connected network, but the shared anycast gateway address is a gateway function, not a separately advertised host endpoint that should appear as ordinary routed host reachability. Treating the anycast IP as a regular host route would undermine the distributed gateway model and create ambiguous ownership across PEs. Reference: IRB anycast gateway, MAC-VRF/IP-VRF interconnection, distributed default-gateway operation.

================

Comprehensive and Detailed 150 to 250 words of Explanation From [SR Linux EVPN and Data Center Interconnect/Course Guide/topics]:

A Gratuitous ARP is an ARP message a host sends to announce or refresh its own IP-to-MAC binding without waiting for another host to request it. In a traditional Ethernet subnet, the GARP is sent as a broadcast so that other hosts can update their ARP caches with the sender's current MAC address. This is useful after a host boots, changes NICs, moves to another attachment point, or takes over an IP address in a redundancy scenario. In EVPN environments, GARPs are also important because a leaf can snoop the ARP information and update local proxy ARP and EVPN MAC/IP state. Option D is false because recipients do not acknowledge a gratuitous ARP with a reply. GARP is an announcement mechanism, not a request/response transaction. If every receiving host acknowledged a broadcast GARP, the result would be unnecessary ARP traffic amplification. The correct behavior is passive update of ARP state by receiving systems and, in EVPN, potential control-plane propagation of the learned binding by the local PE. Reference: GARP behavior, proxy ARP learning, Layer 2 EVPN endpoint update procedures.

================

Comprehensive and Detailed 150 to 250 words of Explanation From [SR Linux EVPN and Data Center Interconnect/Course Guide/topics]:

In all-active Layer 2 EVPN multi-homing, the host is typically dual-homed to two or more leaf routers using a LAG. The participating leaf routers must configure the LAG and associate it with the Ethernet Segment so EVPN can advertise the common ESI and apply aliasing, split-horizon, and DF procedures. If VLAN tagging is used for service separation, tagging must be configured on the LAG interface so that the correct subinterfaces can bind into the MAC-VRF and Ethernet Segment. Option C is false because it states that the LACP system-id-mac must uniquely identify each leaf router. In an all-active EVPN multihomed LAG, the opposite principle applies: from the host's LACP perspective, the multihomed leaf pair must appear as a single logical LACP system. That generally requires a shared LACP system ID or coordinated system MAC behavior across the participating PEs. If each leaf presented a unique LACP system identity, the host would treat them as separate LAG partners and the all-active bundle would not form correctly. Reference: all-active L2 EVPN multi-homing, LAG attachment, LACP system ID behavior, Ethernet Segment association.

================

Comprehensive and Detailed 150 to 250 words of Explanation From [SR Linux EVPN and Data Center Interconnect/Course Guide/topics]:

This scenario describes all-active Layer 2 EVPN multi-homing with a host connected through a LAG to Leaf1 and Leaf2. The LAG subinterface is associated with the MAC-VRF on both participating leaves, and the Ethernet Segment ES-1 is configured for all-active multi-homing. In all-active operation, both leaf routers can be active attachment points for host-originated traffic, and remote traffic can use EVPN multi-homing mechanisms to reach the segment. Option D is false because the host does not know or use the EVPN Designated Forwarder state when sending BUM traffic. The host forwards over its LAG based on its local LAG hashing and LACP behavior. DF election is an EVPN PE-side mechanism used mainly to control which PE forwards BUM traffic from the EVPN overlay toward the Ethernet Segment, preventing duplicate delivery to the multihomed access network. The host itself does not selectively forward all BUM traffic toward the DF. That distinction is critical: DF controls overlay-to-segment replication, while the host's LAG controls host-to-leaf link selection. Reference: all-active L2 EVPN multi-homing, host LAG behavior, DF election scope, BUM forwarding.

================