H3C S12504G-AF Intelligent Core Switches for Data Center Networks

H3C S12504G-AF Intelligent Core Switches for Data Center Networks

H3C’s S12500G-AF switches are part of H3C’s next generation of AI-enabled intelligent switches for core data center use cases, delivering industry-leading switching performance. They have the following features and capabilities:

Multi-level and multi-plane Clos+ switching architecture

High-performance GPUs with over 100 trillion floating point operations per second compute capacity

Industry’s first network-focused AI algorithm

The S12500G-AF series is represented by the S12504G-AF, 12508G-AF and S12516G-AF models, the port density and performance of which allows you to choose a suitable solution for networks of various sizes; they provide a reliable guarantee for creating data center networks and support INT and AD-NET functionality.

Advanced Multi-layer and multi-plane architecture switching Klosa+

Multi-level and multi-plane architecture Klosa+ switching provides the ability to constantly increase throughput, which opens the way for increasing system performance and development, as well as gradually increasing the capacity of the system as a whole.

Thanks Supporting 48-port 10GE, 36-port 40GE/100GE line cards, the switches meet all existing and future data center application needs.

With separation of control planes and data transfer functions, forwarding and control operate in isolation, while the switch matrix slots support redundancy according to the 5+1 or 4+2 scheme.

Redundancy of fan modules and power supplies.

Intelligent networks for account built-in capabilities AI

In the S12500G series switches -AF is supported by the Seerblade High-Performance AI Compute Engine, which provides an intelligent computing platform tightly integrated with networking capabilities, and features high-performance CPU, GPU, large memory, meeting the requirements for simple AI and big data application configurations in small enterprises and mid-sized:

Thanks to the high processing power provided by the high-performance GPU and high-speed network connection, it achieves performance of 123 TFlops (trillion). floating point operations per second), which is a million times higher than the capabilities of traditional processors.

Network-oriented intelligent algorithm, run jointly by upper modules, expands the capabilities intelligent network management and improves performance.

Requirements applications for centers processing< /b> data on database cloud computing

S12500G Series Switches -AF support VXLAN (Virtual eXtensible LAN) technology. VXLAN is a Layer 2 virtual private networking technology over an IP network using MAC address encapsulation within UDP frames. VXLAN technology allows Layer 2 interconnection of distributed physical sites using existing service provider networks or corporate IP networks, as well as providing isolation for different tenant organizations.

S12500G-AF series switches support EVPN (Ethernet Virtual Private Network) technology. EVPN is a Layer 2 virtual private networking technology that uses MP-BGP in the control plane to advertise EVPN routing information, and in the data plane, VXLAN encapsulation is used to forward packets.

The S12500G-AF series switches support large Layer 2 network interconnect technology, which can be used to implement Layer 2 connectivity using EVPN+VXLAN technologies, as well as connections between multiple data center sites.

S12500G-AF series switches support Fiber Channel over Ethernet technology – FCoE; FCoE technology is mainly used to solve the problem of heterogeneous solutions in local area networks of cloud data centers and FC storage networks. A solution to this problem is possible by using FCoE technology in conjunction with Ethernet technology with improved convergence support (Converged Enhanced Ethernet (CEE).

Means ensuring high < b>availability for centers processing data

FFDR technology provides BFD and OAM functions that implement fast failover and protocol convergence. Data center level high availability features include:

BFD for VRRP/BGP/IS-IS/RIP/OSPF/RSVP/static routing


Separation of control and data transmission planes due to independent control processors and switching matrix modules.

Redundancy according to the 1+1 scheme of control processors

Redundancy according to the scheme N+1 switching matrix modules

Redundancy according to the scheme 1+1 fan modules

Redundancy according to the scheme N+M power supplies

< b>Means ensuring high accessibility on based architecture DRNI

S12500G-AF series switches support DRNI (Distributed Resilient Network Interconnect) technology, which implements connection aggregation across different devices by virtually combining two physical devices into one at the forwarding level, while maintaining control plane independence from each other and providing active-active access to the device. This technology provides redundancy at the hardware level and load sharing for traffic processing while increasing the reliability of the system as a whole.

Multi-level means support < b>Security

The S12500G-AF series switches use Quality of Service (QoS) policies to filter and restrict traffic from the data plane to the control plane. In the event of a denial of service attack, the switch is able to detect and protect critical packets, discarding attack-related packets and maintaining normal operation.

The switch supports a large number of access control lists (ACLs) without compromising media-speed forwarding. Access control lists allow you to detect and control L2/IPv4/IPv6/MPLS traffic using various combinations of packet fields

Comprehensive tools maintenance and monitoring

Real-time health monitoring – the presence of a dedicated processor module for monitoring the status of switch matrix modules, backplane channels, communication channels of various services, main chips and memory. If a fault is detected, a corresponding message is sent to the system via the built-in EMS service subsystem

Card Isolation – the specified cards are isolated from the forwarding plane. At the same time, isolated cards continue to operate in the control plane, which allows the user to perform various management operations on them, such as real-time diagnostics and CPLD updates on isolated cards without affecting the functioning of the entire system

OAM functions for Ethernet – provide support for various Device-level and network-level fault detection methods

Energy Saving

Intelligent Processor Maintenance Subsystem (EMS) – Provides intelligent power management and supports sequential power-on and power-off module power supply, as well as monitoring the status of devices. Sequential (alternating) power cycling reduces spikes in power supply and electromagnetic radiation levels, increasing device life. In addition, the device health monitoring function allows you to isolate faulty and idle cards to reduce power consumption

Intelligent fan control – provides temperature information, calculates fan speed and transmits calculated speed information to fan modules. In addition, the system records fan speeds, fault alarms and adjusts fan speeds depending on configurations and areas, reducing power consumption and noise levels, as well as extending fan life

Internal interface monitoring – automatically shuts down unused internal interfaces to reduce energy consumption.

Additional information

Weight 110 kg
Dimensions 26.4 × 44 × 85.7 cm
Switching capacity

57.6 / 387 Tbit/s


21600 million packets/s

MPU module slots


LPU module slots


Высота над уровнем моря при работе


Switch Fabric Module Slots


MPU module name


MPU Module Processor

Quad-core.2.2 GHz

MPU module SDRAM memory

16 GB

MPU module flash memory

8 GB

Console ports on MPU


Management ports (MGMT) on MPU


USB ports on MPU



Redundancy of MPU modules, switch fabric modules, power supplies and fan modules



The page is under development

This section is under development.

We apologize for the temporary inconvenience

Let call
Please enable JavaScript in your browser to complete this form.