H12-921-V1-0 Huawei Practice Questions

This is the fundamental architectural difference. A Type 1 hypervisor (e.g., VMware ESXi, Huawei FusionCompute) is a bare-metal hypervisor that installs directly onto the physical server hardware, acting as its own lightweight OS. A Type 2 hypervisor (e.g., VMware Workstation, Oracle VirtualBox) is a hosted hypervisor that runs as a standard application within a conventional host operating system (like Windows or Linux). Type 1 hypervisors generally offer better performance and are standard for data center environments.

The correct orchestration flow is as follows:

1. The user sends a request to Nova (OpenStack Compute) to create a VM.
2. Nova receives the request and, as part of the provisioning process, makes an API call to Neutron (OpenStack Networking) to create a network port for the VM.
3. Neutron processes the request and, through its plugin/driver, communicates with the external SDN controller, iMaster NCE-Fabric.
4. NCE-Fabric then configures the corresponding physical leaf switch, applying the necessary VLAN/VXLAN settings and security policies for that port.

For a VTEP (leaf switch) to correctly process traffic from a connected host, it must map the host's traffic (identified by its VLAN) to the correct VXLAN Network Identifier (VNI). If the access port for the new server is not correctly associated with the VNI for that L2 segment, the switch will not encapsulate the server's traffic into the VXLAN tunnel. This means the ARP request will never be forwarded across the fabric to other hosts, resulting in the observed behavior.

The problem describes a situation where the reactive mechanism (PFC) is being triggered too often. This indicates that the proactive mechanism (ECN, as part of DCQCN) is not acting quickly enough. By lowering the ECN marking threshold on the switch queues, packets will be marked as congestion-experienced sooner (i.e., at a lower queue depth). This causes the receiving NICs to send CNPs back to the senders earlier, prompting them to reduce their sending rate before the switch buffer fills to the point of triggering a PFC pause. This proactive rate reduction alleviates congestion and reduces reliance on the disruptive PFC mechanism.

The EVPN Type-2 route advertisement is designed to be flexible. It contains fields for both an IPv4 address and an IPv6 address, along with the MAC address. In a dual-stack environment, when a leaf switch learns about a host, it will populate both the IPv4 and IPv6 address fields in a single Type-2 route advertisement. This allows for efficient advertisement of both address families associated with a single endpoint.

In a BGP EVPN VXLAN fabric, border leaf switches function as the gateways between the VXLAN overlay network and external networks. They are responsible for route leaking between tenant VRFs and the public or underlay network. When advertising external prefixes (like those learned from a PE router) into the EVPN fabric, the border leaf originates EVPN Type-5 (IP Prefix) routes. This allows workloads within the fabric to communicate with external services.

DCQCN is a proactive congestion control mechanism that relies on two key actions. First, the switch detects incipient congestion (e.g., buffer queue exceeding a threshold) and marks the ECN bits on passing packets. Second, the destination host's NIC receives these marked packets and sends a Congestion Notification Packet (CNP) back to the original sender, instructing it to reduce its transmission rate. PFC is a reactive mechanism (pausing traffic) that works alongside DCQCN.

Service Function Chaining (SFC) is the specific technology designed for steering traffic through an ordered sequence of service functions, such as firewalls, load balancers, or intrusion detection systems. Within iMaster NCE-Fabric, an administrator would define a service chain that specifies the virtual firewall as a required hop for traffic flowing between the Web, App, and DB tiers (EPGs). This ensures security policies are applied without complex manual routing.

Using eBGP for the underlay is a modern best practice for large-scale spine-leaf fabrics. Assigning a unique private ASN to each leaf switch (or rack) and another to the spine layer creates a simple, highly scalable, and loop-free routing domain. This design avoids the complexities of IGP flooding, simplifies troubleshooting, and aligns well with the BGP-based EVPN overlay. It offers excellent ECMP support for load balancing across the fabric.

FabricInsight's Proactive Path Detection feature allows operators to define an application or flow (based on source/destination IP, port, etc.) and then uses telemetry data to trace and visualize its complete path through the network. This includes the source VM, physical leaf, spine, destination leaf, and destination VM. It highlights latency and packet drop statistics at each hop, enabling rapid identification of performance bottlenecks.