H3C S12500X-AF Data Center Cloud Core Series Switches

Advanced CLOS+ multi-grade multi-plane switching architecture

CLOS+ multi-grade multi-plane architecture, midplane free design, providing continuous bandwidth upgrade capability

Supports industry first 48-port 40GE/ 100GE interfaces and can meet the existing and future application requirements of data centers

Adopts independent switching fabric modules and MPU engines to improve device availability and ensure bandwidth expansion

Virtualization technologies – IRF2

IRF2 can virtualize up to two S12500X-AF switches into one logical IRF fabric. IRF2 delivers the following benefits:

High Availability (HA) - Patented hot standby technology provides data backup and non-stop forwarding on the control plane and data plane. It improves availability and performance, eliminates single-point of failures, and ensures service continuity

Distribution - Multi-chassis link aggregation to enable load sharing and backup over multiple uplinks, improving redundancy and link utilization

Easy Management - A single IP address to manage the whole IRF fabric, which simplifies device and topology management, improving operating efficiency, and lowering network maintenance cost

Virtualization technologies - MDC

MDC virtualizes one S12500X-AF switch into multiple logical switches, enabling multiple services to share one core switch. The 1:N virtualization maximizes switch utilization, reduces network TCO, and ensures secure isolation of services

DC-oriented features

VXLAN (Virtual Extensible LAN) —VXLAN uses a MAC-in-UDP encapsulation method where the original Layer 2 package is added with a VXLAN header, and is then placed in a UDP-IP packet. With the help of MAC-in-UDP encapsulation, VXLAN tunnels Layer 2 network over Layer 3 network which provides two major benefits: higher scalability of Layer 2 segmentation and better utilization of available network paths

MP-BGP EVPN ( Mutliprotocol Border Gateway Protocol Ethernet Virtual Private Network) MP-BGP EVPN uses standard-based BGP protocol as the control plane for VXLAN overlay networks, providing BGP based VTEP auto peer discovery and end-host reachability information distribution. MP-BGP EVPN delivers many benefits, such as eliminating traffic flooding, reducing full mesh requirements between VTEPs via the introduction of BGP RR, achieving optimal flow based end to end load sharing and more

Large capacities for storing ARP/ND, MAC, and ACL entries

RoCE (RDMA over Converged Ethernet)

Remote Direct Memory Access (RDMA) directly transmits the user application data to the storage space of the servers, and uses the network to fast transmit the data from the local system to the storage of the remote system. RDMA eliminates multiple data copying and context switching operations during the transmission process, and reduces the CPU load.

RoCE supports RDMA on standard Ethernet infrastructures. H3C S6850 switch support RoCE and can be used to build a lossless Ethernet network to ensure zero packet loss.

RoCE include the following key features，include PFC(Priority based Flow Control), ECN(Explicit Congestion Notification), DCBX(Data Center Bridging Capability Exchange Protocol), ETS(Enhanced Transmission Selection).

Innovative multi-engine design

Independent control, detection, and maintenance engines provide powerful control capability and millisecond-level HA:

Independent control engine— Uses a powerful CPU system that can efficiently process protocol and control packets, providing refined control for protocol packets and comprehensive protection against protocol packet attacks

Independent detection engine— Provides highly reliable Fast Fault Detection and Restoration (FFDR) such as BFD and OAM, which can interact with protocols on the control plane to implement millisecond-level failover and convergence, ensuring service continuity

Independent maintenance engine—Uses an intelligent Embedded Maintenance Subsystem (EMS), a CPU system that provides smart power management, including sequential power-on and power-off and device status check. Sequential power-on and power-off reduces power impulse, electromagnetic radiation, power consumption, and extends the device lifespan

DC-class HA

FFDR provides BFD and OAM functions to implement fast failover and convergence. The following lists the DC-class HA features:

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

NSR/GR for OSFP/BGP/IS-IS/RSVP

Separation of control and data planes through independent control engine and switching fabric module.

1+1 redundancy for control engines

N+1 redundancy for switch fabric modules

1+1 redundancy for fan trays

N+M redundancy for power modules

Multi-level security protection

The S12500X-AF switch series use QoS policies to filter and limit traffic from data plane to control plane. During a DoS attack, the switch can identify and protect important packets and discard attack packets, ensuring normal operation

Supports a large numbers of ACLs while ensuring line-speed forwarding. ACLs can identify and control L2/IPv4/IPv6/MPLS traffic by using combinations of packet fields

The S12500X-AF switch series supports hardware level encryption technology MACsec (802.1ae), which is an industry-standard security technology that provides secure communication for all traffic on Ethernet links.

Distributed buffering and precise QoS

Distributed ingress buffers accommodate burst traffic. Each port performs a precise bandwidth assignment and traffic shaping for incoming traffic, and distributes the traffic to ingress buffers. Distributed buffering can fully utilize the buffers of line cards to ensure best buffering performance

A network model change from C/S to B/S leads to increased volumes of burst traffic. Network devices must have larger buffering capabilities to support this. The S12500X-AF series supports 200 ms buffering of burst traffic per 10G interface, which can meet the burst traffic requirements of large data centers

Each chip can support 4GB buffer, maximum of 24GB buffer per line card

Each line card supports a maximum of 96K hardware queues, refined QoS, and traffic management. QoS can assign different priorities and queues to different users to provide differentiated services

Comprehensive maintenance and monitoring

Online state monitoring - Uses a dedicated engine to monitor the state of switch fabric modules, backplane channels, service communication channels, key chips, and storage. Once a failure occurs, it reports the failure to the system through EMS

Card isolation- Isolates specified cards from the forwarding plane. The isolated cards still work on the control plane, allowing the user to perform management operations such as real-time diagnosis and CPLD upgrade on the isolated cards without affecting system operation

Ethernet OAM- Provides multiple device-level and network-level fault detection methods

Green

Intelligent EMS engine system - Provides smart power management that supports sequential power-on and power-off and device status check. Sequential power-on and power-off reduces power impulse and electromagnetic radiation, and increases the lifetime of the device. Additionally, device status checks can isolate faulty and idle cards to reduce power consumption

Smart fan management- Collects fan temperature, calculates fan speed, and assigns the calculated speed to the fan tray. In addition, it detects fan speeds, fault alarms, and performs speed adjustment based on configuration sand area, reducing power consumption and noise, increasing the fan's lifetime

Internal interface monitoring-Automatically shuts down unused internal interfaces to reduce power consumption

RoHS compliance - The S12500X-AF switch series meets the EU RoHS safety standards.

The S12500X-AF switch series is designed with front to back air flow, satisfying highly efficient heat dissipation requirements in data center.

Hardware Specifications

Software Specification

Performance and scalability

The typical data center application is an EVPN-VxLAN design,S12500G-AF or S12500X-AF swithes work as spine or spine/border, S68XX series work as leaf and border or ED. From this design ,the users can get a non-blocking large L2 system.