Huawei CloudEngine 6870-48Q6CQ Switch

The Huawei CloudEngine 6870-48Q6CQ is a purpose-built 100GE access-layer switch designed to address the demanding connectivity requirements of AI computing clusters and modern enterprise data centers. It delivers 48 ports of 100GE server-facing connectivity combined with 6 ports of 400GE uplink capacity, enabling a non-blocking, high-throughput fabric architecture that eliminates bottlenecks at the server tier. The switching capacity reaches up to 7.2 Tbps with a forwarding rate of 2,400 Mpps, making it well-suited for east-west traffic patterns characteristic of distributed AI training workloads and large-scale virtualized environments.

At the architecture level, the CE 6870-48Q6CQ supports both VXLAN and SR-MPLS, two cornerstone technologies for building scalable, multi-tenant data center networks. VXLAN enables network virtualization overlays that extend Layer 2 segments across Layer 3 boundaries, critical for workload mobility and micro-segmentation in hybrid cloud and private cloud environments. SR-MPLS support allows traffic engineering with deterministic, low-latency paths without per-flow state maintenance in the network core, making the switch a versatile building block in spine-leaf and multi-tier fabric topologies.

Launched in Q2 2025, the CloudEngine 6870-48Q6CQ incorporates Huawei's latest merchant silicon optimized for high-bandwidth, low-latency operation. The switch supports advanced telemetry including streaming gRPC-based monitoring and In-band Network Telemetry (INT), enabling real-time visibility into queue depths, latency, and link utilization. Its 1RU fixed form factor maximizes rack density, while comprehensive automation capabilities via NETCONF, YANG models, and OpenConfig ensure seamless integration into DevOps-driven infrastructure management pipelines.

Ideal for

• AI and GPU cluster interconnect: Connecting high-density GPU servers at 100GE per node with 400G uplinks to spine switches, enabling full bisection bandwidth for distributed deep learning training jobs.
• Hyper-converged infrastructure (HCI) access layer: Providing non-blocking 100GE connectivity for HCI nodes running VMware vSAN or Nutanix, supporting storage and compute traffic on a single converged fabric.
• Multi-tenant private cloud networking: Leveraging VXLAN overlay support to create isolated virtual network segments for multiple tenants within an enterprise private cloud, maintaining separation without dedicated hardware per tenant.
• Enterprise data center spine-leaf fabric: Serving as a high-density leaf node in a spine-leaf architecture, aggregating server traffic and forwarding it to spine switches via 400G uplinks with SR-MPLS traffic engineering.
• High-performance storage area networking: Connecting NVMe-oF and high-throughput all-flash storage arrays at 100GE, delivering the low-latency, high-IOPS connectivity required by latency-sensitive database and analytics workloads.
• Telemetry-driven network operations: Deploying streaming telemetry and INT capabilities to feed real-time network state data into AIOps platforms for proactive congestion detection and automated remediation in large-scale data center environments.

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.