Marvell

Teralynx® 10 - Data Center Ethernet Switch

Teralynx® 10
The world's lowest-latency programmable switch family with advanced telemetry – 25.6 Tbps through 51.2 Tbps
Teralynx® 10 - Data Center Ethernet Switch

 
Product Description

Marvell Teralynx 10 switches are the world's highest performance switch chips, with up to 51.2 Tbps throughput and industry-leading low latency, analytics and programmability. It is ideal for data center fabrics as well as AI/ML and HPC environments. These switches integrate Marvell's proven, robust 112G LR SerDes, supporting up to 64 ports of 800GbE or Block Diagram 2 x1/10G PCIe3 x4 Secure Control 128 ports of 400GbE. Teralynx 10 enables high radix switch architectures that can collapse data center switch tiers, dramatically reducing cost and power.

Application Field
 
  1. Data center networking infrastructure for the cloud, AI/ML, HPC, high-end enterprise and service providers
  2. Next-generation switches for ToR, leaf, spine and data center interconnect

Features
 
Features Benefits
Switch capacity up to 51.2 Tbps with large buffers Enables customers to deploy fewer network switches and tiers dramatically reducing cost, power, latency, while significantly simplifying management
Up to 512 100G SerDes that support 25G, 50G and 100G I/O speed with proven, robust interoperability
  • Lowers the cost per bit and enables higher scale I/O such as 400 GbE and 800 GbE while maintaining backward compatibility with 25G NRZ
  • OEM & ODM switches can provide high port count 100G, 200G, 400G and 800G connectivity using 50G/100G PAM4 with support for 25/50/100/ 200/400 and 800 GbE connectivity
Industry-leading low latency in cut-through & store-andforward modes Reduces time in networking and shortens job completion time, especially for distributed workloads like AI/ML, HPC and distributed databases
Teralynx® Flashlight™ visibility and telemetry, including support for P4 in-band network telemetry (INT) along with critical extensions Enable predictive, faster and more accurate issue resolution, higher automation, and self-aware, and self-healing networks
Line-rate permutable flex-forwarding to accommodate future networking protocols with software upgrades Enables support for custom and new packet forwarding protocols without requiring ASIC spins 
Power-efficient architecture Facilitates overall data center efficiency and lower total cost of ownership
  • Comprehensive IP forwarding, including congestion aware routing 
  • Highly scalable/flexible layer 2 and 3 tables for IPv4, IPv6 and hybrid  networks
  • Tunneling protocols including IP-in-IP, GRE, MPLS, VXLAN and Geneve
  • Optimized packet forwarding enables higher network utilization and prevents congestion
  • Same switch can be deployed in multiple different use cases, saving cost
Advanced QoS/traffic management feature set including DCB and RoCE
  • Enables easy deployment of applications with stringent network requirements 
  • Delivers higher performance for demanding applications like AI/ML and storage
PTP (IEEE 1588v2) and SyncE support Provides timing and synchronization for applications that are sensitive to timing accuracy and precision (e.g., trading, online gaming, telecom)


*Marvell Authorized Distributor
banner_ins
訂閱電子報,掌握最新科技與產業趨勢
訂閱電子報,掌握最新科技與產業趨勢
我要訂閱

數字驗證

請由小到大,依序點擊數字

洽詢車

你的洽詢車總計 0 件產品

    搜尋

    偵測到您已關閉Cookie,為提供最佳體驗,建議您使用Cookie瀏覽本網站以便使用本站各項功能

    本網站使用Cookie為您提供最佳的使用體驗。繼續使用本網站,即表示您同意我們的Cookie Policy