Mellanox, MT28928A0-NCCF-GE, ConnectX-6 2 x 50GbE Ethernet Network Adapter
Features:
- 10/25/40/50/100 Gb/s Ethernet, PAM4/NRZ
- Up to 100 Gb/s bandwidth
- Message rate of up to 215 Mbps
- Sub 0.8 uses latency
- Mellanox Multi-Host with advanced QoS
- ASAP2
- Accelerated Switching and Packet Processing for virtual switches/routers
- Supports a variety of overlay tunnels
- Network service chaining acceleration for SDN and NFV
- Stateful rule checking for connection tracking
- IPsec and TLS in-line crypto acceleration
- Block crypto acceleration for data-at-rest
- Hardware Root-of-Trust and secure firmware update
- Advanced RoCE capabilities
- Best in class PTP for TSN applications
- Host chaining technology for economical rack design
- Platform agnostic: x86, Power, Arm
Solutions:
- Enterprise data-centers
- Cloud-native, Web 2.0, hyperscale
- Cyber security
- Big data analytics
- Scale-out compute and storage infrastructure
- Telco and Network Function Virtualization (NFV)
- Cloud storage
- Machine Learning (ML) and Artificial Intelligence (AI)
- Low latency, high-frequency financial trading
- High-performance computing (HPC)
- Media and Entertainment
Mellanox® ConnectX®-6 Dx SmartNIC IC is the industry's most intelligent and secure cloud Ethernet network adapter (NIC) ASIC, accelerating mission-critical cloud applications, such as security, virtualization, SDN/NFV, big data, machine learning, and storage, and delivering the highest return on investment (ROI). ConnectX-6 Dx is a member of the award-winning ConnectX series of network adapters whose novel capabilities accelerate cloud and data-center payloads with unprecedented world-class performance and resiliency. ConnectX-6 Dx adapters are equipped with either dual ports of 10/25/40/50 GbE, delivering sub-800 ns latency and up to 215 million messages per second.
Advanced Virtualization
ConnectX-6 Dx delivers another level of innovation to enable building highly efficient virtualized cloud data centers:
- Virtualization - Mellanox ASAP2
- Accelerated Switch and Packet Processing® technology for vSwitch/vRouter hardware offloads deliver orders of magnitude higher performance vs. software-based solutions. ConnectX-6 Dx ASAP2 offers both SR-IOV and VirtIO in-hardware offload capabilities and supports up to 8 million rules.
- Advanced Quality of Service - Includes traffic shaping and classification-based data policing
Security From Zero Trust to Hero Trust
In an era where privacy of information is key and zero trust is the rule, ConnectX-6 Dx adapters offer a range of advanced built-in capabilities that bring security down to the endpoints with unprecedented performance and scalability, including:
- Crypto - IPsec and TLS data-in-motion inline encryption and decryption offload, and AES-XTS block-level data-at-rest encryption and decryption offload.
- Probes & DoS Attacks Protection - ConnectX-6 Dx enables a hardware-based L4 firewall by ASAP2 offloading of stateful connection tracking.
- NIC Security - Hardware Root-of-Trust (RoT) Secure Boot and secure firmware update using RSA cryptography, and cloning-protection, via a device-unique secret key.
Industry-Leading Roce
Following the Mellanox ConnectX tradition of industry-leading RoCE capabilities, ConnectX-6 Dx adds another layer of innovation to enable more scalable, resilient, and easy-to-deploy RoCE solutions.
- Zero Touch RoCE - Simplifying RoCE deployments, ConnectX-6 Dx allows RoCE payloads to run seamlessly on existing networks without requiring special configuration on the network (no PFC, no ECN). The features in ConnectX-6 Dx ensure resiliency and efficiency at the scale of such deployments.
- Configurable Congestion Control - API to build user-defined congestion control algorithms, best serving various environments and RoCE and TCP/IP traffic patterns.
Machine Learning and Big Data
Machine learning applications are based on training a deep neural network, which requires complex computations and fast and efficient data delivery. Mellanox solutions enable smart offloading such as RDMA, GPUDirect®, and more advanced capabilities that dramatically improve neural network training performance and overall machine learning applications. GPUDirect was developed to dramatically improve GPU-to-GPU communication, reducing latency and increasing performance and CPU utilization.