|
Company Name：Kino Technology Limited
Website：www.kino86.com
Address ：Room 312, Honghua Building, No. 1 Guangyayuan Road, Bantian Street, Longgang District, Shenzhen city, Guangdong Province, China
Contact Person：kiki
Tel ：+8613040881925
E-mail：kiki@szkiki.com
Wechat:+8613040881925
Whatspp: +8613040881925
Teams:kiki@szkiki.com
|
|
|
| Product >> Huawei >> ALL |
| |
|
Product_Id： |
87932316 |
ProductName： |
XH16804 |
Specification： |
|
Product Notes： |
|
Product Category： |
Huawei |
| |
|
| Product Description |
Huawei CloudEngine XH16804 Switch Chassis
Product Overview
The Huawei CloudEngine XH16804 is a next‑generation modular spine‑core chassis switch belonging to the XH16800 series, engineered for ultra‑large‑scale AI computing clusters and hyperscale cloud data centers. It adopts advanced backplane‑free Clos cell‑switching architecture with completely separated control plane, fabric forwarding plane and monitoring plane. Equipped with dedicated AI acceleration chips, it delivers enhanced iLossless‑X lossless network capability. The chassis offers 4 hot‑swappable service line‑card slots, supporting high‑density 200GE / 400GE / 800GE high‑speed interfaces. All key hardware components support hot‑swap and multi‑dimension redundancy, providing carrier‑grade stability for mission‑critical large‑model training and inference services.
Key Features
-
Massive non‑blocking switching performance: Designed for the AI era, supports various high‑speed line cards, delivering abundant line‑rate 200GE/400GE/800GE ports to meet extreme bandwidth requirements of spine‑layer in huge AI fabric networks.
-
Enhanced AI‑optimized lossless networking: Dedicated on‑board AI chip runs iLossless‑X algorithm, performing real‑time perception and adaptive optimization for end‑to‑end cluster traffic. Cooperates with PFC, AI‑ECN, NSLB and RoCE‑v2, realizing near‑zero packet‑loss and microsecond‑level latency for RDMA AI workloads.
-
Full‑feature data‑center protocol stack: Hardware‑accelerated VXLAN bridging & routing, BGP‑EVPN, CSS2 cluster virtualization, VS device virtualization, M‑LAG, IEEE 1588v2 PTP, hardware‑based BFD, DPFR and MACsec encryption. Supports large‑scale leaf‑spine fabric construction and cross‑data‑center DCI interconnection.
-
Intelligent real‑time O&M capability: High‑precision millisecond‑grade native Telemetry, IFIT, NetStream and intelligent fault prediction. It implements automatic anomaly identification and fast fault isolation, lowering operation‑maintenance complexity for massive‑node AI clusters.
-
Carrier‑grade modular redundancy design: 1+1 hot‑swappable main‑control units, N+M redundant fabric modules, N+1 hot‑swap power modules and fan trays. Compatible with AC, ‑48 V DC and 240‑380 V HVDC power input. Optimized front‑to‑back airflow to guarantee reliable heat dissipation under high‑load AI traffic scenarios.
Technical Specifications
-
Form factor: 10U standard 19‑inch rack‑mount chassis
-
Service line‑card slots: 4 hot‑swappable LPU slots
-
Fabric module slots: Multiple hot‑swap SFU slots with N+M redundancy
-
Main‑control units: 2 slots for 1+1 hot‑swap backup
-
Management interfaces: RJ45 Console port, RJ45 ETH management port, USB port
-
Switching architecture: Backplane‑free Clos cell‑switching, VoQ virtual output queuing
-
Airflow mode: Fixed front‑to‑back airflow
-
Power modules: Multiple hot‑swappable power slots, support N+1 redundancy; compatible with AC / ‑48 V DC / HVDC
-
Fan trays: Hot‑swappable fan trays with N+1 redundancy
-
Software platform: VRP8 advanced version for AI data center
Application Scenarios
Primarily deployed as high‑end spine core switches in hyperscale AI computing data centers, ultra‑large‑scale HPC clusters and cloud‑native data centers. It undertakes high‑speed aggregation and forwarding of massive leaf‑switch traffic, provides high‑reliability lossless RoCE networking for large‑model AI training, and supports multi‑tenant VXLAN‑EVPN virtual fabric and cross‑data‑center interconnection.
|
|
|
| Click：28 Entry Time：2026-08-07 【Print】 【Close】 |
|
|
|
|
