H19-402_V1.0 Questions and Answers

In Huawei’s intelligent O & M architecture, NETCONF is a modern network management protocol designed for transactional, reliable, and concurrent-safe configuration management . One of its key features is the lock mechanism , which allows a user or system to lock a configuration datastore (such as candidate or running configuration) to prevent other users or processes from making conflicting changes simultaneously. This ensures consistency, atomic operations, and rollback capability , which are critical in automated data center environments like CloudFabric.

On the other hand, SNMP (Simple Network Management Protocol) is primarily designed for monitoring and basic configuration , but it lacks advanced transaction control features. SNMP does not support configuration locking or transactional commits, meaning multiple users can overwrite each other’s changes, potentially leading to configuration inconsistencies.

Huawei strongly recommends using NETCONF/YANG-based automation (e.g., via iMaster NCE-Fabric) for modern data center operations because it provides structured data modeling, conflict prevention, and programmability , which SNMP cannot offer.

Therefore, the statement is TRUE .

In Huawei data center network design, M-LAG (Multi-Chassis Link Aggregation) is widely used to build loop-free Layer 2 topologies without relying on Spanning Tree Protocol (STP) . This is a key advantage in modern CloudFabric architectures.

Traditional Layer 2 networks depend on STP to prevent loops, but STP introduces drawbacks such as blocked links, slow convergence, and inefficient bandwidth utilization . M-LAG eliminates these issues by allowing two physical switches to operate as a single logical device from the perspective of downstream nodes.

With M-LAG:

All links are active-active , maximizing bandwidth usage

Loop prevention is achieved through protocol mechanisms and peer-link synchronization , not STP

Convergence is much faster compared to STP-based designs

Network design becomes simpler and more predictable

Huawei best practices strongly recommend disabling STP in M-LAG-based data center fabrics , especially when combined with VXLAN EVPN, to achieve high performance, fast convergence, and simplified operations .

Therefore, the statement is TRUE .

In Huawei CloudFabric and modern data center architectures, the underlay and overlay networks are designed as independent layers with clear separation of responsibilities. The underlay network provides basic IP connectivity using standard Layer 3 routing (commonly using protocols such as OSPF, IS-IS, or even static routing), ensuring reliable transport between network devices.

The overlay network, typically implemented using VXLAN with BGP EVPN as the control plane, builds logical Layer 2/Layer 3 services on top of the underlay. Importantly, the overlay does not require the same routing protocol as the underlay. This independence allows flexible design, scalability, and easier evolution of services without modifying the physical infrastructure.

Option A is correct because endpoints (servers, VAS devices) are unaware of the underlay—they only interact with the overlay. Option C is correct as VXLAN is the mainstream overlay encapsulation technology. Option D correctly defines the overlay as a logical abstraction over the physical network.

Thus, statement B is false because Huawei explicitly supports different protocols between underlay and overlay layers.

In Huawei data center networking, especially within Intelligent Lossless Data Center Network design, incast traffic is a well-known challenge in distributed systems such as AI computing clusters and distributed storage architectures .

In an incast scenario , a large number of servers simultaneously send traffic to a single destination (e.g., parameter server in AI training or storage node in distributed storage). This sudden aggregation of traffic can quickly overflow switch buffers , leading to packet loss, increased latency, and throughput degradation .

Huawei identifies incast as a primary cause of network congestion in modern data centers. To mitigate this, CloudFabric solutions implement advanced congestion control technologies such as:

PFC (Priority Flow Control) to prevent packet loss

ECN (Explicit Congestion Notification) for early congestion signaling

AI-driven traffic scheduling and load balancing

These mechanisms help build a lossless Ethernet fabric , which is essential for RDMA-based AI workloads and high-performance storage systems.

Therefore, the statement is TRUE , as incast traffic is indeed prone to causing congestion in such environments.

In Huawei CloudFabric VXLAN architecture, distributed gateway networking is designed to overcome the limitations of centralized gateway models by pushing Layer 3 gateway functionality down to leaf nodes .

Options B, C, and D are all true advantages of distributed gateways. B highlights deployment flexibility, where a leaf can act as both Layer 2 and Layer 3 VXLAN gateway, enabling efficient east-west traffic forwarding. C and D emphasize improved scalability: each leaf only maintains local ARP/MAC entries , avoiding the bottleneck of a centralized device learning all endpoints. This significantly reduces resource consumption and enhances horizontal scaling.

Option A , however, describes a characteristic of centralized gateway architecture , where inter-subnet traffic is handled by a central node. This approach leads to suboptimal forwarding paths and potential performance bottlenecks, which distributed gateways are specifically designed to eliminate.

Huawei best practices strongly recommend distributed VXLAN gateways for modern data centers due to their scalability, optimal traffic paths, and reduced resource pressure .

Therefore, the correct answer is A .

In Huawei CloudFabric architecture, the underlay network is defined as a pure IP transport (bearer) network whose primary role is to provide Layer 3 reachability between VTEPs (VXLAN Tunnel Endpoints) . It is designed to be simple, stable, and highly reliable , typically composed of spine and leaf switches running routing protocols such as OSPF, IS-IS, or EBGP .

The statement is false because it incorrectly includes LBs (load balancers) and firewalls as part of the underlay network. In Huawei design principles:

Underlay = physical IP forwarding infrastructure (switch fabric only)

Overlay/services = VAS devices (firewalls, load balancers, etc.)

Firewalls and load balancers are considered service devices (VAS) and are usually connected via service leaf nodes or border leaf nodes , operating in the overlay/service layer , not as part of the underlay forwarding plane.

Additionally, the underlay is typically implemented using a spine-leaf architecture , not traditional three-tier (core/aggregation/access) designs.

Therefore, the statement is FALSE .

RoCEv2 (RDMA over Converged Ethernet version 2) is not based on TCP ; instead, it is implemented over UDP/IP . Therefore, the statement is false.

RoCEv2 encapsulates RDMA traffic directly into UDP packets over IP networks , enabling it to operate across Layer 3 networks, unlike RoCEv1 which is limited to Layer 2. This design allows RoCEv2 to leverage existing IP routing infrastructure in modern data centers.

However, because RoCEv2 does not use TCP, it does not inherently provide congestion control or retransmission mechanisms like TCP does. As a result, Huawei emphasizes the need for a lossless network fabric to support RoCEv2 traffic. Technologies such as:

PFC (Priority Flow Control)

ECN (Explicit Congestion Notification)

DCQCN (Data Center Quantized Congestion Notification)

are implemented in Huawei CloudFabric to ensure reliable and low-latency communication.

This architecture is critical for AI computing, HPC, and distributed storage , where RDMA enables high throughput and ultra-low latency.

Therefore, the correct answer is B (FALSE) .

Huawei data center design documents clearly state that overlay networks (typically based on VXLAN EVPN) are highly versatile and support multiple application scenarios across modern and traditional environments.

In private cloud scenarios , overlay networks enable resource pooling and seamless data center interconnection, allowing flexible workload migration and multi-tenant isolation. For traditional IDC hosting services , overlays provide improved scalability and tenant isolation compared to VLAN-based designs, making them suitable for legacy-to-cloud evolution.

In public cloud environments (IaaS/PaaS/SaaS) , overlay networks are essential for large-scale multi-tenancy, enabling logical network segmentation, automation, and elastic service provisioning. Similarly, in NFV cloud scenarios , overlays integrate with SDN and NFV architectures to support virtualized network functions, service chaining, and dynamic service deployment.

Huawei CloudFabric emphasizes that overlay technology is not limited to a single use case but is a foundational technology across all modern data center architectures. Therefore, all listed options are valid application scenarios.

Huawei’s AI Fabric solution is designed to build a unified, intelligent lossless data center network by integrating traditionally separate network infrastructures into a single converged Ethernet-based fabric.

The three integrated networks are:

Computing network (C): Handles east-west traffic between compute nodes (e.g., AI training clusters, GPU servers).

Storage network (B): Supports high-throughput, low-latency storage access (e.g., distributed storage systems like HDFS or Ceph).

Ethernet network (D): Provides the underlying IP-based transport, enabling unified connectivity and scalability.

Traditionally, InfiniBand (IB) networks were used for high-performance computing due to their low latency and RDMA capabilities. However, Huawei AI Fabric replaces IB with lossless Ethernet (RoCE-based) solutions, allowing all services (compute + storage + management) to run over a single Ethernet fabric .

This convergence reduces:

Network complexity

Capital and operational costs

Management overhead

While improving:

Resource utilization

Scalability

Automation

Therefore, the correct three networks are Storage, Computing, and Ethernet , making B, C, and D correct .

In Huawei CloudFabric architecture, logical switches are a key abstraction used to simplify network service deployment and management. These logical switches are built on top of VXLAN Layer 2 networks and represent virtualized broadcast domains (similar to VLANs, but more scalable).

Huawei extends the function of these logical switches by integrating distributed Layer 3 gateway capabilities directly into them. This means that each logical switch can provide both Layer 2 connectivity and Layer 3 routing services (commonly referred to as distributed gateways). These gateways are typically implemented on leaf nodes using VXLAN with BGP EVPN as the control plane.

As a result, traffic between subnets can be routed locally at the ingress leaf switch without needing to traverse to a centralized gateway, significantly improving efficiency and reducing latency. Logical ports connected to these logical switches automatically receive gateway services, enabling seamless inter-subnet communication.

This design aligns with Huawei’s intent of decoupling physical and logical networks while providing scalable, agile, and highly efficient data center networking. Therefore, the statement is TRUE.

In Huawei CloudFabric and VXLAN-based overlay networking, the overlay provides abstraction, flexibility, and scalability by decoupling logical services from the physical infrastructure. NVEs (Network Virtualization Edge nodes) can indeed be deployed on physical switches (such as leaf switches), which is a fundamental capability of hardware-based VXLAN fabrics. Additionally, overlay networks are designed to meet the needs of private cloud environments, offering high performance, simplified O & M, and enhanced security through network segmentation (e.g., VXLAN VNIs).

Overlay networks also support access for both virtual machines and physical servers, ensuring consistent connectivity regardless of workload type.

However, statement D is incorrect. In Huawei’s architecture, especially with BGP EVPN-based VXLAN, the control plane is distributed rather than relying on a controller to synchronously deliver flow tables. Devices dynamically learn and exchange forwarding information using EVPN. Even in SDN-based scenarios with iMaster NCE-Fabric, policy delivery is not strictly synchronous flow table pushing in the traditional sense (like OpenFlow).

Therefore, “controller delivers flow tables synchronously” is not a standard capability of Huawei overlay networks, making option D incorrect.

RDMA (Remote Direct Memory Access) is a key technology in Huawei’s Intelligent Lossless Data Center Network , widely used in high-performance scenarios such as AI training, distributed storage, and HPC.

The statement is correct because RDMA enables direct memory-to-memory data transfer between servers without involving the CPU or operating system in the data path. This bypass mechanism significantly reduces overhead and enables:

Ultra-low latency (microsecond-level communication)

High throughput (efficient use of bandwidth)

Low CPU utilization , freeing compute resources for applications

Although RDMA was originally developed for InfiniBand networks , it is now widely used over Ethernet through technologies like RoCE (RDMA over Converged Ethernet) in Huawei CloudFabric solutions.

To support RDMA effectively, Huawei implements lossless Ethernet mechanisms such as PFC, ECN, and congestion control algorithms to ensure reliable packet delivery.

Therefore, the statement accurately describes RDMA characteristics and is TRUE .

In Huawei CloudFabric architecture, iMaster NCE-Fabric acts as the central network controller and automation platform that directly integrates with server virtualization platforms such as FusionSphere, VMware vCenter, and OpenStack . This integration enables automated network provisioning, policy enforcement, and dynamic resource orchestration between compute and network layers.

Through northbound APIs, iMaster NCE-Fabric communicates with virtualization platforms to obtain tenant, VM, and network information. It then translates these requirements into fabric configurations, such as VXLAN, EVPN, and underlay routing policies. This tight integration is essential for automated service deployment and end-to-end orchestration , which is a core capability in Huawei’s CloudFabric solution.

On the other hand, iMaster NCE-FabricInsight is focused on intelligent O & M, monitoring, and analytics . It provides telemetry-based visibility, fault detection, and performance optimization, but it does not directly connect to or integrate with virtualization platforms for service provisioning.

Therefore, the correct answer is iMaster NCE-Fabric , as it is the component responsible for connecting to server virtualization systems and enabling automation.