Free H12-821_V1.0 Exam Questions – 2025 Updated

The sequence of events follows the "Physical Layer $\rightarrow$ BFD Protocol $\rightarrow$ Upper-Layer Application" causality chain:

1. Link fault alarm: When the physical link is disconnected, the physical layer interface immediately detects the loss of signal and reports a physical link fault (Link Down). This is the root trigger.
2. The BFD session goes down.: BFD is linked to the interface status. Once the physical interface reports a fault (or BFD stops receiving packets), the BFD local session state transitions to Down.
3. Interrupted OSPF neighbor relationship: Upon transitioning to the Down state, the BFD process notifies the upper-layer application (OSPF). OSPF reacts to this notification by immediately tearing down the neighbor adjacency, without waiting for the standard Hello/Dead timers to expire.

Huawei Technologies Co., Ltd., "BFD Configuration Guide - BFD for OSPF," Huawei VRP Documentation. This document details the interaction flow: "The BFD session quickly detects a link fault and notifies OSPF... OSPF then quickly responds to the network topology change."

Katz, D., & Ward, D. (2010). Bidirectional Forwarding Detection (BFD). IETF RFC 5880. Section 6.8.6. (Describes the transition to the Down state upon detection of a failure).

Huawei Support, "Overview of BFD," Huawei Enterprise Support Community. (Describes the "Hardware detection" mechanism where link faults are detected and propagated to BFD).

1. pat: The question requires configuring NAPT (Network Address Port Translation). In Huawei USG firewall syntax, the command mode pat enables NAPT, which allows multiple private IPs to share a single public IP by translating ports. (Using no-pat would result in a one-to-one mapping, which is not NAPT).
2. trust: The source-zone parameter defines where the traffic originates. As shown in the topology diagram, PC1 is located in the trust zone (12.1.1.0/24 network) and is initiating traffic toward the untrust zone.
3. 12.1.1.0: The source-address command specifies the packet source filter. The syntax shown includes a trailing 24, indicating a subnet mask length. The value 12.1.1.0 correctly identifies the subnet 12.1.1.0/24 where PC1 resides.
4. inner: The command action source-nat address-group binds the NAT policy to a specific address pool. The configuration snippet at the top creates a NAT address group named inner ([FW1]nat address-group inner), so the policy must reference this exact name to function.

Huawei Technologies Co., Ltd. (n.d.). Configuring NAT Rules in a NAT Policy. Huawei Support - USG6000 Series Documentation. Retrieved from https://info.support.huawei.com

Huawei Technologies Co., Ltd. (n.d.). CLI: Example for Configuring NAT No-PAT. Huawei Support - Security Configuration Guide. (Verifies mode pat vs mode no-pat syntax and source-zone usage).

Huawei Technologies Co., Ltd. (n.d.). Nat-policy command reference. USG6000 Series Command Reference. (Verifies source-address syntax requiring network ID when mask is specified).

The process of assigning a terminal (STA) to a VLAN via a VLAN pool follows a logical "Access -> Lookup -> Assignment -> Utilization" workflow:

1. Association (Access): The process initiates when the STA associates with the Virtual Access Point (VAP).
2. Configuration Check (Lookup): The Access Controller (AC) or AP identifies the specific VLAN pool configured for that VAP and determines which allocation algorithm (e.g., hash-based or sequential) is active.
3. Allocation (Assignment): The system executes the specified algorithm to select a specific VLAN ID from the pool and assigns it to the STA.
4. Operation (Utilization): Once the assignment is complete, the STA uses the tagged VLAN to transmit data and access network resources.

Huawei Technologies Co., Ltd. AirEngine Wireless Access Controller (AC) Configuration Guide, "Configuring a VLAN Pool." [Official Vendor Documentation].

Relevant Section: "Implementation of VLAN Pool... When a STA goes online, the AC selects a VLAN from the VLAN pool based on the configured algorithm and assigns the VLAN to the STA."

Huawei Technologies Co., Ltd. HCIA-WLAN V3.0 Training Materials, "Chapter: WLAN Service Configuration - VLAN Pool."

Context: Describes the workflow: STA association -> VAP profile lookup -> Algorithm execution -> VLAN assignment.

Huawei Enterprise Support. "Example for Configuring a VLAN Pool." Campus Switch and WLAN Product Documentation.

Detail: Confirms the sequence of determining the bound pool/algorithm upon VAP access before assignment occurs.

In the Virtual Router Redundancy Protocol (VRRP), the Master router is responsible for forwarding packets sent to the virtual router's IP address. To facilitate this, the Master must respond to ARP requests for the virtual IP with the associated virtual MAC address and forward any frames arriving at that virtual MAC. Conversely, a Backup router monitors the Master's status but must remain passive in the data plane to prevent duplicate traffic or IP conflicts. Therefore, a Backup router must not respond to ARP requests for the virtual IP and must discard any packets destined for the virtual MAC address (unless it transitions to the Master state).

IETF RFC 3768, "Virtual Router Redundancy Protocol (VRRP)", Section 6.4.2 (Backup): States that the Backup router "MUST NOT respond to ARP requests for the IP address(es) associated with the virtual router" and "MUST discard packets with a link layer destination MAC address equal to the virtual router MAC address."

IETF RFC 3768, "Virtual Router Redundancy Protocol (VRRP)", Section 6.4.3 (Master): States that the Master router "MUST respond to ARP requests for the IP address(es) associated with the virtual router" and "MUST forward packets with a link layer destination MAC address equal to the virtual router MAC address."

Cisco Systems, "First Hop Redundancy Protocols Configuration Guide, Cisco IOS Release 15M&T", Chapter: VRRP, Section: "VRRP Operation".

Juniper Networks, "Junos OS High Availability Library for Routing Devices", Section: Understanding VRRP, Subsection: "VRRP Router Priority and Preemption".

IGMP (Internet Group Management Protocol): This is a host-router signaling protocol. Its primary function is to manage group membership on the last-hop subnet (the segment connecting the router to the hosts). Hosts use it to join groups, and routers use it to query for members.

PIM (Protocol Independent Multicast): This is a Layer 3 routing protocol used between routers. It builds distribution trees (Shortest Path Trees or Shared Trees) to route multicast traffic across the network infrastructure from the source to the receiver's last-hop router.

IGMP Snooping: This is a Layer 2 optimization mechanism running on switches. Without it, switches broadcast multicast traffic to all ports (flooding). IGMP Snooping "snoops" on IGMP messages to build a forwarding table, ensuring traffic is only forwarded to ports with interested receivers, thus suppressing flooding.

IETF RFC 2236, Internet Group Management Protocol, Version 2, Section 1 "Introduction". (Defines IGMP's role in allowing hosts to report group memberships to immediate neighbor routers).

https://doi.org/10.17487/RFC2236

IETF RFC 7761, Protocol Independent Multicast - Sparse Mode (PIM-SM): Protocol Specification (Revised), Section 1 "Introduction". (Describes PIM's role in building distribution trees to forward multicast packets based on routing information).

https://doi.org/10.17487/RFC7761

IETF RFC 4541, Considerations for Internet Group Management Protocol (IGMP) and Multicast Listener Discovery (MLD) Snooping Switches, Section 1 "Introduction". (Explains how snooping switches limit the flooding of multicast traffic on Layer 2).

https://doi.org/10.17487/RFC4541

Cisco Systems, IP Multicast Technology Overview, "Multicast Basics". (Confirming the functional distinction between IGMP for member management, PIM for routing, and Snooping for L2 constraint).

Document ID: 29232

The Root Port election on a non-root bridge follows a specific four-step tie-breaking sequence to identify the best path to the Root Bridge.

1. RPC Comparison: The process begins by comparing the Root Path Cost (accumulated cost to the root). The port with the lowest cost is preferred.
2. Peer BID Comparison: If path costs are equal, the switch compares the Sender (Peer) Bridge ID. The port connected to the upstream switch with the lowest Bridge ID is selected.
3. Peer PID Comparison: If the Sender Bridge IDs are identical (indicating parallel links to the same switch), the switch compares the Sender (Peer) Port ID. The port receiving frames from the upstream port with the lowest PID is preferred.
4. Local PID Comparison: Finally, if all previous criteria are tied, the Local Port ID is compared, and the lowest value is chosen.

IEEE Standard 802.1D-2004, Media Access Control (MAC) Bridges. Section 17.6 "Port Role Selection", Paragraph 17.6.2. (The standard details the vector comparison: {RootID, RootPathCost, BridgeID, PortID}). https://ieeexplore.ieee.org/document/1339632

Huawei HCIA-Datacom V1.0 Training Material, Spanning Tree Protocol. Section: "STP Calculation - Root Port Election". (Explicitly lists the order: Root Path Cost -> Sender BID -> Sender PID -> Local PID).

Juniper Networks Documentation, Understanding Spanning Tree Protocol Port Selection. (Confirms the standard industry algorithm: Cost -> Sender Bridge ID -> Sender Port ID -> Receiver Port ID).

OSPF defaults its network type based on the Layer 2 (Link Layer) protocol of the interface:

1. Broadcast: This is the default for multi-access, broadcast-capable media like Ethernet and FDDI.
2. P2P (Point-to-Point): This is the default for serial interfaces running PPP (Point-to-Point Protocol) or HDLC. PPP is the standard answer for P2P links in this context.
3. NBMA (Non-Broadcast Multi-Access): This is the default for non-broadcast WAN technologies like Frame Relay, X.25, and ATM.
4. P2MP (Point-to-Multipoint): There is no protocol that defaults to P2MP. It is a manually configured type used specifically on NBMA media (Frame Relay) to resolve reachability issues (hub-and-spoke topologies) without requiring a full mesh or DR/BDR election. Therefore, it maps to Frame Relay.

Huawei Technologies Co., Ltd., "Configuring Network Types for OSPF Interfaces," Huawei Enterprise Support, VRP Documentation. (States: "The network type of Ethernet interfaces is broadcast... If the link layer protocol is PPP... defaults to P2P... If the link layer protocol is Frame Relay... defaults to NBMA.")

IETF, "RFC 2328 - OSPF Version 2," Section 8.1. (Defines the default network types for Broadcast, Point-to-Point, and NBMA networks).

H3C, "OSPF Configuration Guide," H3C Support. (Confirms P2MP is forcibly changed from other types, typically on NBMA networks like Frame Relay).

NetworkAcademy.io, "OSPF Network Types," (Explains the relationship between WAN technologies like Frame Relay and the configuration of P2MP mode).

In Huawei modular device architecture, the planes are physically separated to ensure reliability and performance.

• SFU (Switch Fabric Unit): Functions as the data plane core. It is a switching matrix that connects all LPUs, allowing data to pass from one module to another (inter-card communication) at high speeds.
• LPU (Line Processing Unit): Handles the physical layer and forwarding duties. It houses the optical/electrical ports where cables connect and performs line-rate packet forwarding.
• MPU (Main Processing Unit): Centralizes the control and management planes. It runs the operating system, calculates routing tables (RIB), and manages the entire chassis configuration and monitoring.

Huawei Technologies Co., Ltd. (2024). HCIA-Datacom V1.0 Training Material, "Module 3: Network Engineering - Device Hardware Architecture," pp. 7-12.

Huawei Technologies Co., Ltd. (2023). CloudEngine 12800 Series Switches Hardware Description, "Hardware Structure - MPU, LPU, and SFU Functions."

Huawei Technologies Co., Ltd. (2024). NetEngine 8000 Series Routers Hardware Description, "System Architecture - Control, Data, and Monitoring Planes."

The statement is accurate. Both Modular QoS Command-Line Interface (MQC) and Policy-Based Routing (PBR) are frameworks applied to device interfaces to manage and control traffic.

Filtering Packets: Both MQC and PBR utilize traffic classifiers, which are commonly defined using Access Control Lists (ACLs). These classifiers match packets based on various criteria (e.g., source/destination IP, protocol, port number). This process of matching specific traffic to apply a policy is a form of filtering.

Controlling Forwarding Paths: PBR's fundamental purpose is to override the default destination-based routing table and force packets to take a specific path (e.g., to a different next-hop or out a specific interface). MQC can also be used to control forwarding paths because PBR is implemented within the MQC framework. A redirect action, which defines the PBR forwarding behavior, is configured within a traffic behavior and then applied to an interface using a traffic policy.

Therefore, both mechanisms are applied on interfaces and can be used to filter traffic (through classification) and control forwarding paths.

A "False" answer would be incorrect because it overlooks two key facts. First, the traffic classification step, which is essential for both MQC and PBR, functions as a filtering mechanism to select which packets the policy applies to. Second, it ignores that PBR is a specific function that can be configured as an action within the more general MQC framework, explicitly allowing MQC to control packet forwarding.

1. Huawei HedEx Documentation for AR Routers - IP Routing Configuration Guide - PBR: In the "PBR Overview" section

it states

"PBR is implemented using a traffic policy. The traffic policy is applied to an interface to control the forwarding of packets received on the interface." It further details that a traffic classifier (using if-match acl) is used to match packets

and a traffic behavior (using redirect) is used to specify the forwarding action. This confirms PBR uses filtering (classification) and controls paths on an interface.

2. Huawei HedEx Documentation for AR Routers - QoS Configuration Guide - MQC: The "MQC Overview" section describes the MQC model

consisting of a traffic classifier

a traffic behavior

and a traffic policy. In the "Configuring a Traffic Behavior" section

the redirect command is listed as a possible action. The command description explicitly states its function is to "Configure PBR in a traffic behavior." This directly links MQC's capabilities to implementing PBR for path control.

3. Huawei HedEx Documentation for CloudEngine Switches - PBR Configuration: The documentation states

"PBR allows you to set your own routing policies based on your requirements... PBR is implemented through traffic policies. A traffic policy consists of traffic classifiers and traffic behaviors." This reinforces that the core mechanism for PBR involves classification (filtering) and behavior (path control) and is applied via a policy

consistent with the MQC structure.

The statement accurately describes the default behavior for BGP router ID selection on Huawei devices running the Versatile Routing Platform (VRP). When a BGP process is initiated, the system requires a unique 32-bit router ID to identify the BGP speaker. The selection follows a specific hierarchy:

1. The router ID explicitly configured using the router-id command in the BGP view is used first.

2. If no router ID is explicitly configured, the system selects the highest numerical IP address among all configured and active loopback interfaces.

3. If no loopback interfaces are configured, the system then selects the highest numerical IP address among all configured and active physical interfaces.

The scenario in the question—no explicit router ID and multiple loopback interfaces—directly corresponds to the second step in this selection process. Therefore, BGP will select the largest (highest) loopback interface address as its router ID.

This is a "TRUE OR FALSE" question. The alternative, "False," is incorrect because the described behavior is the standard, documented default mechanism for BGP router ID selection in Huawei's VRP operating system under the specified conditions.

1. Huawei HedEx Documentation (for NetEngine Routers): BGP Configuration Guide > Configuring Basic BGP Functions > (Optional) Configuring a BGP Router ID. This section explicitly states: "If no router ID is specified for a BGP process

the device selects a router ID for the BGP process in the following order: 1. The device preferentially selects the largest IP address among all loopback interface addresses as the router ID. 2. If no loopback interface is configured

the device selects the largest IP address among all interface addresses as the router ID."

2. Huawei HCIP-Datacom-Core Technology V1.0 Training Material: Chapter 4: BGP Fundamentals > Section: BGP Router ID. The training material details the router ID selection principle

confirming that in the absence of a manually configured ID

the highest IP address of a loopback interface is preferred. If multiple loopback interfaces exist

the one with the highest IP address is chosen.

The statement accurately describes the role and connectivity requirements of a non-client in a BGP Route Reflector (RR) topology. By definition, a non-client is an Internal BGP (IBGP) peer that is not configured as a Route Reflector or a client of an RR. The route reflection rules are designed to relax the full-mesh requirement for clients, but not for non-clients.

Specifically, a Route Reflector propagates routes as follows:

A route learned from a client is reflected to all other clients and all non-clients.

A route learned from a non-client is reflected only to clients.

A route learned from an EBGP peer is reflected to all clients and non-clients.

Because an RR does not reflect routes learned from one non-client to another non-client, the original IBGP split-horizon rule remains in effect between them. To ensure complete routing information exchange and prevent routing loops, all non-clients must maintain a full-mesh peering relationship with the Route Reflector(s) and with all other non-clients in the Autonomous System.

This is a "TRUE OR FALSE" question. The provided statement is factually correct based on the principles of BGP Route Reflection. Therefore, the "False" option is incorrect.

1. Huawei HedEx Documentation (VRP): In the BGP Configuration Guide for Huawei VRP-based products

the section on Route Reflection details the roles and rules. It states

"An RR reflects routes learned from a non-client peer only to its client peers." This rule necessitates that non-clients maintain a full mesh among themselves to exchange routes.

Source: Huawei VRP V800R012C10 Configuration Guide - BGP

Section: "BGP Route Reflector".

2. RFC 4456 - BGP Route Reflection: An Alternative to Full Mesh Internal BGP: This is the foundational standard that defines the behavior of Route Reflectors.

Reference: Section 4

"Operation"

describes the route advertisement rules. It specifies that if a route is received from a "non-Client-peer"

the RR should "reflect to all the Clients". It makes no provision for reflecting these routes to other non-Client-peers

thus upholding the full-mesh requirement between them.

DOI: https://doi.org/10.17487/RFC4456

3. University Courseware: Networking courses covering advanced routing protocols explain this principle. For example

materials on BGP scalability solutions would cover Route Reflectors.

Source: Stanford University

CS244: Advanced Topics in Networking. Lecture notes on "Interdomain Routing with BGP" discuss scalability issues and solutions like Route Reflectors

explaining that the full-mesh requirement is only relaxed for clients

not for peers outside the client group (non-clients).

The statement accurately describes the function of the BGP ConnectRetry timer within the BGP Finite State Machine (FSM). The primary purpose of this timer is to control the interval between attempts to establish a TCP session with a BGP peer.

1. On TCP Connection Success: When a BGP speaker successfully establishes a TCP connection with its peer, it transitions from the Connect or Active state to the OpenSent state. In this new state, the router's focus shifts to exchanging BGP OPEN messages and monitoring the session with the Hold Timer. The ConnectRetry timer, having fulfilled its purpose of retrying the TCP connection, is stopped or disabled for this established session.

2. On TCP Connection Failure: If the TCP connection attempt fails, the BGP speaker transitions to the Active state. In this state, it waits for the ConnectRetry timer to expire. Upon expiration, the router initiates a new TCP connection attempt by transitioning back to the Connect state and restarting the timer. This cycle continues until a TCP connection is successfully established or the BGP process is manually shut down.

The "False" option is incorrect because the provided statement correctly summarizes the behavior of the ConnectRetry timer as defined by the BGP protocol standard and implemented by Huawei. The timer is integral to the process of both retrying failed TCP connections and ceasing those retries once a connection is successfully formed.

1. Huawei HCIP-Datacom-Core Technology V1.0 Training Material.

Reference: In the "BGP Fundamentals" chapter

the section on the "BGP State Machine" describes the transitions related to the ConnectRetry timer.

Details: The material explains that in the Active state

"If the ConnectRetry timer expires

BGP changes its state to Connect" to re-attempt the TCP connection. It also clarifies that upon a successful TCP connection

the state changes to OpenSent

at which point the ConnectRetry timer's function is complete and it is stopped.

2. RFC 4271 - A Border Gateway Protocol 4 (BGP-4).

Reference: Section 8.2.2

"Finite State Machine".

Details: This foundational document for BGP outlines the state transitions. For the Active state

it specifies: "If the ConnectRetryTimerExpires event occurs

the local system restarts the ConnectRetry timer

initiates a TCP connection to the other BGP peer... and changes its state to Connect." For a successful TCP connection (TcpConnectionValid or TCPConnectionConfirmed event)

the state transitions to OpenSent

and the description of subsequent actions focuses on the Hold Timer

indicating the ConnectRetry timer is no longer active for the session.

The statement is correct. The global BGP routing table contains over a million prefixes, and processing, storing, and transmitting this volume of information consumes significant router resources (CPU and memory). To manage this burden and to enforce specific routing policies, BGP provides robust route filtering mechanisms. Network administrators use these mechanisms, such as route policies applied in an outbound direction, to control precisely which routes are advertised to BGP peers. This ensures that only necessary routes are propagated, which is a fundamental practice for maintaining network stability, conserving resources, and implementing business agreements between autonomous systems.

This is a "TRUE OR FALSE" question. The provided statement is factually correct and describes a standard BGP operational principle and best practice. Therefore, the "False" option is incorrect.

1. Huawei HCIP-Datacom-Core Technology V1.0 Training Material: In the chapter "BGP Route Control"

the section on "Using a Route-Policy to Filter Routes to Be Advertised" explicitly details this concept. It states

"To implement routing policies and reduce the number of routes on a network

you can use a route-policy to filter routes to be advertised." This directly supports the statement that filtering is necessary to manage the large number of routes and that it can be configured to advertise only specific routes.

2. Huawei VRP 8 Product Documentation - BGP Configuration Guide: The section describing BGP route policies and their application explains the use of the peer {group-name | ipv4-address} route-policy export command. The documentation clarifies that this command applies a routing policy to filter routes advertised to a BGP peer or peer group. This confirms the technical implementation for the practice described in the question.

3. Kurose

J. F.

& Ross

K. W. (2021). Computer Networking: A Top-Down Approach (8th ed.). Pearson. In Chapter 5

"The Network Layer: Control Plane

" the section on BGP discusses inter-AS routing policies. The text explains that an AS can control the routes it advertises to its neighbors. For example

a stub AS will typically only advertise prefixes for its own network

not routes learned from its provider. This filtering is a core aspect of BGP policy enforcement and scalability

aligning perfectly with the question's premise.

The statement is false. A mismatch in OSPF network types between two directly connected routers will prevent them from forming a full neighbor adjacency and synchronizing their Link-State Databases (LSDBs). OSPF network types must be compatible for a neighbor relationship to be successfully established.

A Point-to-Point (P2P) interface and a Point-to-Multipoint (P2MP) interface have fundamentally different operational behaviors and expectations for how OSPF Hello packets are processed and how the link topology is perceived. This incompatibility prevents the OSPF neighbor state machine from progressing to the Full state. Even if parameters like the Area ID and Hello/Dead intervals are identical, the network type mismatch is a critical failure point that will cause the routers to get stuck in a state prior to LSDB synchronization (typically Init or 2-Way). Therefore, making the Hello intervals the same is a necessary but insufficient condition for adjacency formation in this scenario.

The "True" option is incorrect because it overlooks the fundamental requirement for compatible network types in OSPF. The network type dictates crucial aspects of OSPF operation on an interface, including packet formats and the neighbor state machine logic. An incompatibility here cannot be overcome simply by aligning timers.

1. Huawei HedEx Documentation (for VRP V800R012C10): In the OSPF Configuration Guide

the section on establishing OSPF neighbor relationships details the necessary conditions. While it lists matching parameters like Area ID

authentication

and timers

it also describes the distinct behaviors of each network type. The documentation implicitly confirms incompatibility by describing separate operational models. For instance

P2P networks do not elect a Designated Router (DR)

while other types might. The logic for processing Hello packets differs

which is the root of the incompatibility.

Source: Huawei VRP V800R012C10 Product Documentation

"IP Routing > OSPF Configuration Guide > OSPF Fundamentals > Establishment of an OSPF Neighbor Relationship".

Reference Detail: This section outlines the strict conditions for adjacency. The operational differences between P2P and P2MP

as described in the "Network Types" subsection

prevent the state machine from completing its progression.

2. RFC 2328 - OSPF Version 2: This is the foundational standard for OSPF. The RFC defines the behavior for each network type in separate sections

highlighting their distinct operational characteristics.

Source: IETF RFC 2328

"OSPF Version 2".

Reference Detail: Section 9.1

"The Router-LSA

" describes how the link data is advertised differently based on the network type (e.g.

a "point-to-point connection to another router" for P2P vs. multiple "point-to-point connection to a neighbor" entries for P2MP). Section 10

"The Hello Protocol

" details the state machine. A P2P interface expects to see its own Router ID in the neighbor's Hello to transition to 2-Way

but the packet processing and neighbor list handling are incompatible with a P2MP interface

causing the state transition to fail.

3. Computer Networking: A Top-Down Approach by J. Kurose and K. Ross: Reputable networking textbooks used in university curricula explain that for OSPF routers to become adjacent

they must agree on several interface parameters.

Source: Kurose

J. F.

& Ross

K. W. (2021). Computer Networking: A Top-Down Approach (8th ed.). Pearson.

Reference Detail: Chapter 5

"The Network Layer: Control Plane." The chapter's discussion on OSPF explains that routers exchange Hello packets to establish adjacencies and that parameters on the link interface

including the network type's operational mode

must be compatible for this process to succeed. A mismatch leads to a failure to form an adjacency.

The statement is false because it oversimplifies the network requirements for AI workloads by identifying low latency as the sole core requirement. While low latency is a critical factor, it is one of three equally important core requirements for AI data center networks. Efficient AI training, especially large-scale distributed training, depends on a network fabric that provides high bandwidth, low latency, and zero packet loss.

1. High Bandwidth: AI models and datasets are massive, requiring high throughput for efficient data loading and synchronization of model parameters (gradients) between processing units (GPUs/NPUs).

2. Low Latency: Crucial for reducing the idle time of expensive processing units during frequent, small data exchanges in tightly-coupled parallel computations.

3. Zero Packet Loss: This is arguably the most critical requirement for many AI workloads. Packet loss triggers TCP retransmission mechanisms, which introduce significant delays and can cause severe performance degradation, effectively stalling the entire distributed training job. Technologies like RDMA over Converged Ethernet (RoCEv2), which are standard in AI clusters, rely on a lossless network fabric to achieve their high-performance potential.

Therefore, the core requirement is a combination of these three characteristics, not just low latency.

This is a "TRUE OR FALSE" question. The statement is categorized as False because it presents an incomplete and therefore inaccurate conclusion. While the premise about the AI era focusing on data is correct, the conclusion that this leads only to a core requirement of low latency is a logical flaw.

1. Official Vendor Documentation:

Huawei

"AI Fabric Intelligent and Lossless Data Center Network White Paper." In the section discussing network challenges for AI services

it is explicitly stated that AI training clusters require a network with high bandwidth

low latency

and zero packet loss. The paper emphasizes that "the most important thing is to build a zero-packet-loss network" because packet loss in a high-speed network can lead to a sharp decrease in throughput and a significant increase in latency

severely impacting AI training efficiency. (This white paper is a primary source for Huawei's perspective on AI networking).

2. Peer-reviewed Academic Publications:

Chen

J.

et al. (2018). "A Survey on Network-Accelerated Distributed Deep Learning." IEEE Communications Surveys & Tutorials

21(4)

pp. 3494-3521. In Section III-A

"Communication Requirements

" the authors state: "Distributed deep learning systems have three main communication requirements: high bandwidth

low latency

and high scalability... In addition

for RDMA-based communication

a lossless network is required to avoid performance degradation caused by packet loss." This confirms that low latency is just one of several key requirements. DOI: https://doi.org/10.1109/COMST.2018.2875922

3. University Courseware:

MIT 6.888/18.337J

"Parallel Computing and Scientific Machine Learning." Course materials often discuss the communication overhead in distributed computing. Lectures on parallelizing deep learning models highlight that the communication network is a primary bottleneck and its performance is a function of bandwidth

latency

and network topology. The need for lossless fabrics when using protocols like RDMA is a key topic in ensuring efficient scaling of training jobs across multiple nodes. (e.g.

discussions on AllReduce communication patterns and their sensitivity to network performance).

In OSPF, authentication can be configured at two primary levels: the area level and the interface level. The principle of configuration precedence dictates that a more specific configuration overrides a more general one. Therefore, an authentication mode configured directly on an OSPF-enabled interface will always take priority over the authentication mode configured for the entire area to which that interface belongs. If an interface does not have a specific authentication configuration, it will inherit the settings from its parent area. The statement that interface authentication is preferentially used is correct because it has a higher priority.

The alternative, "False," is incorrect. If "False" were true, it would imply that area-level authentication overrides interface-level authentication, or that there is no defined priority. This contradicts the hierarchical configuration logic used in OSPF on Huawei devices and in the protocol standard, where specific settings are designed to provide granular control over general settings.

1. Huawei HedEx Documentation (for VRP - Versatile Routing Platform):

Document: NE40E & NE9000 V800R010C10SPC500 Configuration Guide - OSPF

Section: OSPF Configuration > Configuring OSPF Authentication

Reference Text: "If authentication is configured in an OSPF area and on an interface in this area

the authentication configuration on the interface takes precedence. If no authentication is configured on the interface

the interface inherits the authentication configuration of the OSPF area." This statement explicitly confirms the priority of interface-level authentication.

2. Huawei HCIP-Datacom-Core Technology V1.0 Training Material:

Document: HCIA-Datacom V1.0 Training Material (Foundational concepts are carried over to HCIP)

Chapter: OSPF Basics > OSPF Packet Authentication

Reference Text: The training material explains that OSPF authentication can be set per-area using the area authentication-mode command and per-interface using the ospf authentication-mode command. It clarifies that the interface-level command provides a more granular configuration and will be used for packets sent out of that specific interface

overriding the area-wide setting.

3. Academic/Standard Protocol Behavior (RFC 2328):

Document: RFC 2328 - OSPF Version 2

Section: Appendix D: Authentication

Reference Text: While the RFC defines the authentication fields and types

the implementation of configuration hierarchy (interface vs. area) is a vendor-specific implementation detail. However

the ability to set authentication parameters on a per-interface basis (as described in Section D.3

"The Authentication key") implies the necessity for such a hierarchy. Vendor documentation

like Huawei's

clarifies this implementation-specific precedence rule

which aligns with industry-standard network design principles.

The display bgp routing-table command output provides detailed information about BGP routes. The Path/Ogn field displays the ASPath and the Origin attribute. In the provided image, this field shows 100 i. The i is the Origin code, which stands for IGP. This code signifies that the route was advertised into BGP on the originating router using the network command. This is the preferred origin type in BGP route selection.

A. The Path/Ogn field shows the ASPath as 100, meaning the route was learned from a peer in AS 100, not AS 200.

B. The output explicitly lists the MED (Multi-Exit Discriminator) value for this route as 0, not 100.

C. The Status field shows >. The > symbol specifically indicates that this route is the optimal (best) route selected by the BGP process.

1. Huawei HCIP-Datacom-Core Technology V1.0 Training Material: In the BGP Path Attributes section

the Origin attribute is described. It states

"Origin is a mandatory

well-known attribute that defines the origin of routing information. It has three types: IGP (i): The route is generated using the network command... Incomplete (?): The route is imported using the import-route command." This directly supports that 'i' corresponds to the network command.

2. Huawei HedEx VRP V800R011C10 Command Reference: The documentation for the display bgp routing-table command explains the output fields.

Status: "The second symbol indicates whether the route is the optimal route. > indicates the optimal route." This confirms that the route in the question is optimal

making option C incorrect.

Path/Ogn: "ASPath/Origin attribute of the route. The origin attribute defines the origin of a BGP route. It has the following types: i: indicates that the route is an IGP route." This confirms the meaning of the origin code 'i'.

The Virtual Router Redundancy Protocol (VRRP) uses a standardized virtual MAC address format. This format is 00-00-5E-00-01-XX, where the first five octets are fixed, and the last octet (XX) is the hexadecimal representation of the Virtual Router ID (VRID).

In this scenario, the VRID is 3. The hexadecimal equivalent of the decimal number 3 is 03. Therefore, by substituting the VRID into the standard format, the resulting virtual MAC address is 00-00-5E-00-01-03. The virtual IP address is not used in the calculation of the virtual MAC address.

A. This option uses an incorrect prefix (01-00-5E) and incorrectly uses the last octet of the IP address (64 for 100) instead of the VRID.

B. This option uses an incorrect prefix (01-00-5E). The first octet for a VRRP virtual MAC address must be 00.

C. This option incorrectly uses the hexadecimal value of the last octet of the IP address (64 for 100) as the final octet.

1. Huawei. (2021). CloudEngine S12700 V200R019C10 Configuration Guide - IP Unicast Routing. Section: VRRP Principles > Basic Concepts > Virtual MAC Address. The documentation states

"The virtual MAC address is in the format of 0000-5e00-01xx

where xx indicates the VRID."

2. Cooper

D.

et al. (2010). Virtual Router Redundancy Protocol (VRRP) Version 3 for IPv4 and IPv6. RFC 5798. Internet Engineering Task Force (IETF). Section 7.3

"Virtual Router MAC Address". This RFC specifies the IPv4 VRRP MAC address as 00-00-5E-00-01-{VRID}. (DOI: https://doi.org/10.17487/RFC5798)

3. Purdue University. (n.d.). ECE 46300: Introduction to Computer Networks - Lecture 16: Routing. Slide 31

"VRRP". The course material illustrates the VRRP MAC address format as 00:00:5e:00:01:XX

where XX is the group ID (VRID).

A Type 3 LSA, also known as a Summary-LSA, is generated by an Area Border Router (ABR) to advertise routes learned from one area to other areas. The Link State ID (LSID) field within a Type 3 LSA contains the IP network address of the destination being advertised, along with its subnet mask. The Router ID of the ABR that originates this LSA is contained in the "Advertising Router" field of the LSA header, not in the Link State ID field. Therefore, the statement that the Link State ID indicates the ABR's router ID is false.

A: This statement is true. An OSPF router generates a distinct Type 1 LSA (Router-LSA) for each area to which it belongs, describing the state and cost of its interfaces within that specific area.

B: This statement is true. OSPF enforces a hierarchical, two-level area structure. All inter-area traffic must pass through the backbone area (Area 0). Direct routing information exchange between two non-backbone areas is prohibited.

C: This statement is true. A Type 1 LSA (Router-LSA) describes a router's local links. It defines four types of connections: point-to-point (P2P), transit network (TransNet), stub network, and virtual link.

1. RFC 2328 - OSPF Version 2

Section A.4.3

Page 210.

Regarding Type 3 LSAs (Network-summary-LSAs)

it states: "The Link State ID of this LSA is the network's IP address." This directly refutes option D. The Advertising Router field is defined in the LSA header (Section 12.1) as the Router ID of the LSA's originator.

2. Huawei HedEx Documentation

HCIP-Datacom-Core Technology V1.0

OSPF LSA Types.

The official training material explains that for a Type-3 LSA

the "Link State ID is the address of the destination network segment

" and the "Advertising Router is the router ID of the ABR that generates the LSA." This confirms that statement D is false.

3. RFC 2328 - OSPF Version 2

Section 12.4.1

Page 156.

Regarding Router-LSAs

it states: "A router originates a router-LSA for each area it belongs to." This validates statement A. The same section also details the four link types (point-to-point

transit

stub

virtual)

validating statement C.

4. RFC 2328 - OSPF Version 2

Section 3.1

Page 22.

This section describes the OSPF backbone's role: "The backbone is responsible for distributing routing information between non-backbone areas. The backbone itself is an OSPF area

and all OSPF areas must be connected to the backbone." This validates statement B.

In a Layer 3 roaming scenario, the station (STA) moves to a Foreign AP (FAP) managed by a Foreign AC (FAC) on a different IP subnet. To maintain the STA's original IP address and ensure service continuity, a mobility tunnel is established between the FAC and the Home AC (HAC). Uplink data from the STA is sent to the FAP, which forwards it to the FAC. The FAC then encapsulates this traffic and sends it through the tunnel to the HAC. The HAC decapsulates the traffic and forwards it to the upper-layer network, making it appear as if the STA is still on its home network.

A. This path incorrectly includes the original Home AP (HAP) after the Home AC (HAC) and creates a loop back to the HAC. The STA is no longer associated with the HAP.

C. This path bypasses the Home AC (HAC). This would occur if the STA obtained a new IP address from the foreign network, which contradicts the principle of seamless Layer 3 roaming.

D. This path incorrectly routes traffic from the Home AC (HAC) back to the original Home AP (HAP), which the STA is no longer connected to.

1. Huawei HCIP-WLAN Certification Training V1.0 Materials: In the chapter on "WLAN Roaming Technology

" the section describing Layer 3 roaming details the data forwarding process. It states

"Uplink data packets from the STA are sent from the FAP to the FAC

encapsulated by the FAC

and then sent to the HAC through the mobility tunnel. The HAC then forwards the packets to the upper-layer network." This process directly corresponds to the path described in option B. (Refer to HCIP-WLAN V1.0 Training Material

Chapter: WLAN Roaming Technology

Section: Layer 3 Roaming Principles).

2. Huawei HedEx Documentation for AC6605: The feature guide on WLAN roaming explains the data flow for inter-AC Layer 3 roaming. It illustrates that after roaming

the FAC establishes a CAPWAP tunnel with the HAC to forward the user's data packets

ensuring the user's IP address remains unchanged. The data path shown is STA -> FAP -> FAC -> HAC -> IP Network. (Refer to AC6605 Product Documentation

Feature Description - WLAN

Section: Roaming).

Bidirectional Forwarding Detection (BFD) has two operating modes. Asynchronous mode is the primary and default mode where two systems periodically exchange BFD Control packets. If a system does not receive a packet from its peer within the negotiated detection time, the session is declared down. Demand mode is an alternative where, after a BFD session is established, the systems stop sending periodic BFD Control packets to reduce overhead. Packets are only sent if an explicit connectivity check is required. The Echo function is a supplementary mechanism that can be used alongside Asynchronous mode to test the forwarding path.

B. The BFD Echo function is specifically designed to be used in conjunction with Asynchronous mode; it is not a mutually exclusive feature.

1. IETF RFC 5880: Bidirectional Forwarding Detection (BFD)

D. Katz

D. Ward

June 2010.

Section 6.3

"Asynchronous Mode": "In Asynchronous mode

the systems periodically send BFD Control packets to one another... If a number of packets in a row are not received from the other system

the session is declared to be down." (Supports option C).

Section 6.4

"Demand Mode": "In Demand mode

no BFD Control packets are exchanged after the session is established." (Supports option A).

Section 3.1

"Operating Modes": "Asynchronous mode is the primary mode of operation for BFD..." (Supports option D).

Section 6.5

"Echo Function": "The Echo function may be used in conjunction with Asynchronous BFD Control packet exchange." (Refutes option B).

2. Huawei HedEx Documentation

NE40E & NE5000E V800R011C10 Configuration Guide - Feature Description - BFD.

Section 2.1.2

"BFD Operating Modes": This section describes Asynchronous mode as the most common mode where devices periodically send control packets. It also details Demand mode

where periodic packets are not sent after session setup. This documentation confirms that Asynchronous mode is the primary mode and that it can be used with the Echo function.

The Neighbor Discovery Protocol (NDP) is a core protocol in the IPv6 suite, responsible for interactions between nodes on the same link. Its functions include:

(A) Address resolution: NDP uses Neighbor Solicitation (NS) and Neighbor Advertisement (NA) messages to resolve an IPv6 address to its corresponding link-layer (MAC) address, replacing the function of ARP in IPv4.

(B) Neighbor state tracing: Known as Neighbor Unreachability Detection (NUD), this function tracks the reachability status of neighbors, transitioning them through states like REACHABLE, STALE, and PROBE.

(C) Duplicate Address Detection (DAD): Before an interface uses an IPv6 address, it performs DAD to ensure no other node on the link is using the same address.

(D) Redirection: Routers use Redirect messages to inform hosts of a more optimal first-hop router for a specific destination.

All the provided options are correct and represent key functions of the Neighbor Discovery Protocol.

1. Huawei. (2020). HCIP-Datacom-Core Technology V1.0 Training Material. Section "IPv6 Basics"

Subsection "Neighbor Discovery Protocol (NDP)". This material explicitly lists Address Resolution

Neighbor Unreachability Detection (NUD)

Duplicate Address Detection (DAD)

and Redirection as primary functions of NDP.

2. Narten

T.

Nordmark

E.

Simpson

W.

& Soliman

H. (2007). RFC 4861: Neighbor Discovery for IP version 6 (IPv6). IETF.

Address Resolution is detailed in Section 7.2.

Neighbor Unreachability Detection (state tracing) is detailed in Section 7.3.

Duplicate Address Detection is detailed in Section 5.4.

The Redirect function is detailed in Section 8.

3. Kurose

J. F.

& Ross

K. W. (2017). Computer Networking: A Top-Down Approach (7th ed.). Pearson. Chapter 4

Section 4.4.3 "IPv6"

discusses the ICMPv6 protocol

which includes the functions of NDP such as address resolution and router discovery

corroborating these as standard protocol features taught in university curricula.

An RSTP edge port is designed for interfaces connected to terminal devices, not other switches. A fundamental rule for an edge port is that it should not receive Bridge Protocol Data Units (BPDUs). If an edge port receives a configuration BPDU, the switch interprets this as a connection to another switch, which could create a loop. Consequently, the port immediately loses its edge port status and reverts to a normal RSTP port. It will then participate in the standard RSTP calculation, which involves transitioning through the Discarding and Learning states before potentially reaching the Forwarding state. It does not remain in the Forwarding state.

A. This is a true statement. Edge ports are assumed to connect to end-stations and cannot form loops, so they are excluded from the RSTP topology calculation to ensure rapid convergence.

B. This is a true statement and the primary benefit of an edge port. It bypasses the standard Listening and Learning states, allowing it to transition immediately from Discarding to Forwarding.

D. This is a true statement. Since an edge port connects to a terminal, its state change (up or down) does not alter the core network topology between switches, so it does not trigger a Topology Change Notification (TCN).

1. Huawei HCIP-Datacom-Core Technology V1.0 Training Material:

Section 2.3.2 RSTP: "If an edge port receives a BPDU

it immediately loses the edge port attribute and becomes a common STP port. Then

STP calculation is performed

which may cause network flapping." This confirms that statement C is false. The material also validates statements A

B

and D as correct behaviors of an edge port.

2. Huawei CloudEngine S5732-H Series Switches Product Documentation:

Section: Configuring Basic STP Functions > Configuring an Edge Port: "An edge port does not receive BPDUs. If an edge port receives a BPDU

it immediately loses its edge port attribute and becomes a common STP port. It then participates in STP calculation

causing network flapping." This official vendor documentation directly supports the reasoning that receiving a BPDU causes the port to lose its edge status.

In IS-IS, the router that wins the DIS election on a broadcast LAN originates a pseudonode LSP whose ID format is .-00.

In R1’s Level-2 LSDB the only pseudonode entry is “R3.01-00”, meaning the router with system-ID R3 generated the L2 pseudonode (PN 01). Therefore, R3 is the Level-2 DIS; all other routers have only their normal “00-00” LSPs and no pseudonode LSPs, showing they are not DIS.

1. Huawei NetEngine 8000 Series Configuration Guide – IS-IS

“DIS Election and Pseudonode LSP”

Sec. 8.1

p.152–153.

2. Huawei HCIP-Datacom Courseware: IS-IS Fundamentals

Module 03

slides 24-26.

3. ISO/IEC 10589:2002

Intermediate System to Intermediate System Intra-Domain Routing

§7.3.5 (Pseudonode LSP format)

p.62-63.

4. IEEE Communications Surveys

“Understanding IS-IS Routing”

vol. 20

no. 1

2018

§III-B (DIS election)

p.218.

The ORIGINATORID is an optional, non-transitive BGP attribute created by a Route Reflector (RR) to prevent routing loops. When the first RR (RTB) receives the route 10.1.1.0/24 from its client (RTA), it creates the ORIGINATORID attribute and sets its value to the Router ID of the route's originator, which is RTA's Router ID (1.1.1.1). RTB then reflects this route to its other client, RTC. When RTC, also an RR, receives this route, it sees that the ORIGINATORID attribute already exists. According to the BGP route reflection rules, an RR does not modify an existing ORIGINATORID. Therefore, RTC reflects the route to its client RTD with the ORIGINATORID unchanged. RTD receives the route with the ORIGINATORID of 1.1.1.1.

1. Bates

J.

Chandra

R.

Katz

D.

& Rekhter

Y. (2006). BGP Route Reflection: An Alternative to Full Mesh Internal BGP (iBGP). RFC 4456

Internet Engineering Task Force. Section 8

"Operation"

page 7. Retrieved from https://doi.org/10.17487/RFC4456. This document states

"The RR creates a new attribute

ORIGINATORID

and sets its value to the ROUTERID of the originator of the route. An RR does not create a new ORIGINATORID if one already exists."

2. Huawei. (2021). NE40E-X8 V800R012C10 Configuration Guide - BGP. Section: "Route Reflector". This guide explains

"OriginatorID: ... It is created by an RR and carries the router ID of the route originator. The OriginatorID attribute is used to prevent routing loops."

3. Kurose

J. F.

& Ross

K. W. (2021). Computer Networking: A Top-Down Approach (8th ed.). Pearson. Chapter 5

Section 5.6.3

"BGP Route Selection". The text describes iBGP scaling solutions

including route reflectors and the function of attributes like ORIGINATORID for loop prevention.

On a broadcast LAN, IS-IS elects the router with the highest interface DIS priority. All four routers boot at the same time, so uptime or existing adjacencies do not affect the result. The figure shows R4’s interface priority is the highest among R1–R4; therefore R4 wins the Designated Intermediate System (DIS) election.

1. Huawei CloudEngine Series Switches

IS-IS Configuration Guide V800R010C10

“DIS Election

” p.118-119 – states the router with the highest DIS priority becomes DIS; MAC address is used only when priorities tie.

2. RFC 1195

“Use of OSI IS-IS for Routing in TCP/IP and Dual Environments

” §7.2.7

pp. 33-34 – describes LAN DIS election based first on priority

then on MAC address.

3. MIT 6.829 “Link-State Routing” lecture notes (Fall 2020)

slide 8 – summarises IS-IS DIS selection: highest priority wins

simultaneous start gives equal uptime.

In the PIM multicast routing table an (S,G) entry identifies a source-based tree, also called the Shortest Path Tree (SPT). This tree is built from each receiver back to the specific source S for group G and is used in both PIM-DM and PIM-SM. The alternative (,G) entry represents a shared tree that exists only in PIM-SM. Hence, (S,G) entries are specifically for establishing the SPT on any PIM network.

1. Huawei

“VRP8 PIM Configuration Guide

” Section 3.2.1 “Multicast Routing Entries

” pp. 35-36: “An (S

G) entry represents a source tree/Shortest Path Tree (SPT)…valid for PIM-DM and PIM-SM.”

2. Huawei

“IP Multicast Technology White Paper

” v2.0

§4.1

p. 18: “(S

G) → SPT; (

G) → RP/shared tree. SPT applies to both PIM modes.”

3. RFC 4601 (Protocol Independent Multicast – Sparse Mode)

§2.4: “(S

G) state represents the SPT created for packets from source S to group G.”

The default Access Control List (ACL) matching order on Huawei routers is config. In this mode, the system matches packets against ACL rules sequentially, based on the ascending order of rule IDs (e.g., rule 5, rule 10, rule 15). The matching process ceases once a packet matches a rule, and the corresponding permit or deny action is taken. This "first-match" principle is the standard behavior. An alternative auto order exists, where the system reorders rules based on precision (depth-first), but this must be explicitly configured.

1. Huawei HedEx Documentation: NE40E V800R012C10SPC500 Configuration Guide - ACL and QoS

Section: "ACL Configuration"

Subsection: "Configuring the Matching Order for ACL Rules". The guide states

"By default

an ACL uses the configuration order (config) for matching."

2. Huawei HCIP-Datacom-Core Technology V1.0 Training Material: Chapter 4: ACL Technology

Section: "ACL Rule Matching Order". This official courseware explains the two matching orders

config and auto

and explicitly identifies config as the default order used by the VRP (Versatile Routing Platform) operating system.

3. Huawei Campus Switch S5700 V200R021C00 Product Documentation: Configuration > Security > ACL Configuration Guide > Configuring an ACL > (Optional) Configuring the Matching Order for ACL Rules. This document confirms the default behavior: "The system matches packets against ACL rules in config order by default."

Protocol Independent Multicast (PIM) routers use Hello messages on multicast-enabled interfaces to discover PIM neighbors and maintain adjacencies. This is a fundamental mechanism for PIM operation on a multi-access network segment. According to the PIM-SM specification defined in RFC 7761, and as implemented by default in Huawei's VRP (Versatile Routing Platform), the standard interval for sending these Hello messages is 30 seconds. This value, known as PIMHelloPeriod, ensures timely neighbor discovery without generating excessive control plane traffic. The corresponding neighbor hold time is typically 3.5 times this interval, or 105 seconds.

1. Internet Engineering Task Force (IETF). (2016). RFC 7761: Protocol Independent Multicast - Sparse Mode (PIM-SM). Section 4.3.2

"Hello Message Generation". The document states

"Hello messages are sent periodically to each PIM-enabled interface. The default period is PIMHelloPeriod seconds. The default value for PIMHelloPeriod is 30 seconds." Available: https://doi.org/10.17487/RFC7761

2. Huawei Enterprise Support. (2021). NE40E V800R012C10 Configuration Guide - IP Multicast. Section 5.3.1

"PIM-SM Configuration". The documentation for the pim timer hello command confirms the default setting. It states

"By default

the interval for sending Hello messages is 30 seconds." (Document ID: EDOC1100091898).

3. Kurose

J. F.

& Ross

K. W. (2021). Computer Networking: A Top-Down Approach (8th ed.). Pearson. In Chapter 4

Section 4.8

"IP Multicast: The Distance Vector Multicast Routing Protocol (DVMRP)"

while discussing general multicast principles that extend to PIM

the mechanisms of neighbor discovery through periodic messages are detailed

with PIM's 30-second default being a standard industry example.

In Huawei VRP, the command aggregate-address ip-address mask summary-only generates a single aggregate route and suppresses all component-specific routes, so only the summary prefix is advertised. Because the attributes of the covered routes are not retained, the aggregate carries the Atomic-aggregate attribute instead of inheriting the individual community attributes, exactly as stated in the question.

1. Huawei

“BGP Command Reference

” Release 8

Section “aggregate-address”

p. 128: “summary-only—Suppresses all specific routes and advertises only the summarized route; the route is originated with the Atomic-aggregate attribute.”

2. RFC 4271: Border Gateway Protocol 4

Sec. 5.1.6

p. 22: “If the aggregator omits path information

it MUST add the ATOMICAGGREGATE attribute.”

3. Huawei

“HUAWEI NE40E&80E Product Documentation

” Feature Guide – BGP

Chap. 4.3 “Route Aggregation

” p. 4-32: “Use the summary-only keyword to ensure only the aggregate route is sent while sub-prefixes are suppressed.”