NVIDIA MSN3420-CB2FC Spectrum SN3420 25/100GbE Open Ethernet Switch

The NVIDIA MSN3420-CB2FC is a Spectrum-2 based SN3420 1U Open Ethernet switch pre-installed with NVIDIA Cumulus Linux. It provides 48 ports of 10/25GbE (SFP28) and 12 ports of up to 100GbE (QSFP28), delivering up to 2.4Tb/s throughput and 3.58Bpps processing capacity. Designed as a top-of-rack (ToR) leaf platform, it is ideal for cloud, hyperconverged storage, and data analytics deployments.

The NVIDIA Spectrum SN3000 series switches are based on the 2nd generation of Spectrum switches, purpose-built for leaf/spine/super-spine data center applications. The SN3000 series provides port speeds spanning from 1GbE to 400GbE and a port density that enables full rack connectivity to any server at any speed. Uplink ports allow a variety of blocking ratios to suit application requirements.

The MSN3420-CB2FC is the SN3420 model pre-installed with NVIDIA Cumulus Linux. As data center switching architectures increasingly adopt 100GbE, the SN3420 offers a high-performance, cost-effective way to evolve host connectivity from 10G to 25G. Equipped with 48 ports of 10/25GbE and 12 ports of up to 100GbE in a compact 1U form factor, the SN3420 is an ideal ToR switch platform, delivering a total throughput of up to 2.4Tb/s with a processing capacity of 3.58Bpps. It enables seamless use of QSFP28 connections for leaf-spine topology and future-proofs the data center.

The SN3000 series is based on the high-performance Spectrum-2 ASIC with a bidirectional switching capacity of 6.4Tb/s. It delivers high performance combined with feature-rich Layer 2 and Layer 3 forwarding, full wire speed, cut-through mode latency, on-chip fully-shared 42MB packet buffering, and flexible port use. Combining innovations in programmability, telemetry, and tunneling with industry-leading performance, the SN3000 series addresses today's data center's complex networking requirements.

• 48 ports of 10/25GbE (SFP28) and 12 ports of up to 100GbE (QSFP28)
• Up to 2.4Tb/s throughput and 3.58Bpps processing capacity
• Spectrum-2 ASIC with 6.4Tb/s bidirectional switching capacity
• Pre-installed with NVIDIA Cumulus Linux
• 42MB fully shared on-chip packet buffer
• 425ns cut-through latency
• 1+1 hot-swappable power supplies and N+1 hot-swappable fans
• Reversible airflow option
• Color-coded PSUs and fans
• Multi-chassis LAG for active/active L2 multipathing
• 64-way ECMP routing for load balancing and redundancy
• NVIDIA What Just Happened (WJH) telemetry
• 512K on-chip flow counters
• 512K shared forwarding entries shared across ACL, LPM routes, host routes, MAC, ECMP, and tunnel applications
• Advanced network virtualization with single-pass VXLAN routing and IPv6 segment routing
• Cloud-scale NAT — 100K+ sessions
• Programmable pipeline that can programmatically parse, process, and edit packets
• Deep packet inspection — 512B deep
• RoHS compliant

The Spectrum-2 packet buffer is fully shared across all ports, supporting cut-through line rate traffic from all ports without compromising scale or features. With its fast packet buffer, Spectrum-2 provides a high-performance, fair, and bottleneck-free data path for mission-critical applications. Spectrum-2 features a programmable pipeline and a deep packet parser/editor capable of processing payload up to the first 512B. It supports single-pass VXLAN routing as well as bridging, and advanced virtualization features such as IPv6 segment routing and network address translation (NAT).

Spectrum-2 provides deep and contextual network visibility, enabling network operators to proactively manage issues and reduce mean time to recovery. The What Just Happened (WJH) Spectrum feature leverages underlying silicon and software capability to provide granular, event-triggered information about infrastructure issues. Rich telemetry information from Spectrum-2 is readily available via open APIs that are integratable with third-party software tools and workflow engines.

Spectrum-2 uses intelligent algorithms and efficient resource sharing to support unprecedented forwarding table, counter, and policy scale. Fine-grained resource allocation allows up to 512K entries to be dynamically shared across MAC, ARP, IPv4/IPv6 routes, ACLs, ECMP, and tunnels. An innovative algorithmic TCAM optimized for data centers and cloud environments can scale the number of rules to up to half a million rules.

Open Ethernet breaks the paradigm of traditional switch systems by eliminating vendor lock-in. Instead of forcing network operators to use the specific software provided by the switch vendor, Open Ethernet offers the flexibility to use a choice of operating systems on top of Ethernet switches, thereby regaining control of the network and optimizing utilization, efficiency, and overall return on investment. Encouraging an ecosystem of open source, standard network solutions, Open Ethernet adopts the same principles as standard open solutions for servers and storage and applies them to networking infrastructure.

Cumulus Linux is a powerful open network operating system enabling advanced automation, customization, and scalability using web-scale principles like the world's largest data centers. It accelerates networking functions and provides choice from an extensive list of supported switch models including NVIDIA Spectrum-based switches. Cumulus Linux was built for automation, scalability, and flexibility, allowing you to build data center and campus networks that ideally suit your business needs. Cumulus Linux is the only open network OS that allows you to build affordable and efficient network operations like the world's largest data center operators, unlocking web-scale networking for businesses of all sizes.

Cumulus NetQ is a highly scalable, modern network operations tool set that provides visibility, troubleshooting, and lifecycle management of your open networks in real time. NetQ delivers actionable insights and operational intelligence about the health of your data center and campus networks — from the container or host all the way to the switch and port, enabling a NetDevOps approach. NetQ is the leading network operations tool that utilizes telemetry for deep troubleshooting, visibility, and automated workflows from a single GUI interface, reducing maintenance and network downtimes. With full lifecycle management functionality, NetQ combines the ability to easily upgrade, configure, and deploy network elements with a full suite of operations capabilities, such as visibility, troubleshooting, validation, trace, and comparative look-back functionality.

SN3000 series platforms are available out of the factory in three different flavors: pre-installed with NVIDIA Cumulus Linux; pre-installed with NVIDIA Onyx, a home-grown operating system utilizing common networking user experiences and an industry standard CLI; or bare metal including an ONIE image, installable with any ONIE-mounted OS. ONIE-based platforms utilize the advantages of open networking and the NVIDIA Spectrum-2 ASIC capabilities.

Additional operating systems supported across the SN3000 series include SONIC, Linux Switch, and DENT. SONIC was designed for cloud networking scenarios where simplicity and managing at scale are the highest priority. On the SN3000 series, SONIC enables fine-grained failure recovery and in-service upgrades (ISSU) with zero downtime. Linux Switch enables users to natively install and use any standard Linux distribution as the switch operating system, such as DENT, a Linux-based networking OS stack suitable for campus and remote networking.

The Open Network Install Environment (ONIE) is an Open Compute Project open source initiative driven by a community to define an open install environment for bare metal network switches such as the NVIDIA SN3000 series. ONIE enables a bare metal network switch ecosystem where end users have a choice of different network operating systems. NVIDIA also fully supports the running of third-party containerized applications on the switch system itself, with the third-party application having complete access to the bare-metal switch via direct access to the SDK. The switch has tight controls over the amount of memory and CPU cycles each container is allowed to use, along with fine-grained monitoring of those resources.

• Top-of-rack (ToR) leaf switching in cloud and enterprise data centers
• Web-scale IT, cloud, and hyperconverged storage environments
• Data analytics clusters requiring high throughput and consistent low latency
• Leaf-spine topologies migrating from 10G to 25G host connectivity
• Layer 2 and Layer 3 cloud-scale networks with VXLAN and IPv6 segment routing
• RoCE transport for NVMe over Fabric and machine learning applications using GPUDirect
• NetDevOps environments using Cumulus Linux and Cumulus NetQ for automation and telemetry

The SN3420 supports a broad range of transceivers and cables across 10GbE, 25GbE, 40GbE, 50GbE, and 100GbE, including DAC, AOC, SR4, PSM4, LR4, CWDM4, SWDM4, ER, and DR1 options. It also supports 100GbE to 4x 25GbE breakout DAC and AOC, and 100GbE to 2x 50GbE breakout DAC and AOC, as well as QSA pluggable adapters for 40GbE to 10GbE and 100GbE to 25GbE conversion.

The SN3000 series is part of the NVIDIA complete end-to-end solution, which provides 1GbE through 400GbE interconnectivity within the data center. Other devices in this solution include ConnectX-based network interface cards and LinkX copper or fiber cabling. Because system-level compatibility can depend on the operating system, firmware, and transceiver/cable combination, please confirm the exact configuration before ordering.

• Cost-effective 25G evolution: Provides a high-performance path to evolve host connectivity from 10G to 25G with 100GbE uplinks.
• Open Ethernet flexibility: Eliminates vendor lock-in and supports multiple operating systems, including Cumulus Linux, Onyx, SONIC, Linux Switch, and DENT.
• Industry-leading performance: 2.4Tb/s throughput and 3.58Bpps processing capacity with 425ns cut-through latency.
• Pervasive visibility: What Just Happened telemetry, 512K on-chip flow counters, and streaming telemetry reduce mean time to issue resolution.
• Massive scale: 512K shared forwarding entries flexibly shared across ACL, LPM routes, host routes, MAC, ECMP, and tunnels.
• Advanced virtualization: Single-pass VXLAN routing, IPv6 segment routing, and cloud-scale NAT with 100K+ sessions.
• High availability: 1+1 redundant power supplies, N+1 redundant fans, multi-chassis LAG, and 64-way ECMP.
• End-to-end solution: Part of the NVIDIA complete data center solution with ConnectX NICs and LinkX cabling.

Starsurge Group provides pre-sales configuration support, compatibility checks, and post-sales technical assistance for NVIDIA networking products. Our team can help confirm whether the MSN3420-CB2FC is suitable for your ToR, leaf-spine, or storage fabric design, and can advise on alternative SN3000 series models, operating systems, or airflow configurations when needed.

NVIDIA SN3000 series switches come with a one-year limited hardware return-and-repair warranty, with a 14 business day turnaround after the unit is received. Support services including next business day and 4-hour technician dispatch are available. NVIDIA offers installation, configuration, troubleshooting and monitoring services, available on-site or remotely delivered.

• Confirm the required airflow direction: this part number is P2C (port-to-connector) airflow.
• Confirm the required operating system: this SKU is pre-installed with Cumulus Linux. Choose another SKU if Onyx or ONIE is required.
• Verify rack space and depth requirements: 1U form factor, 17" depth.
• Verify power and cooling capacity: 2 power supplies, 5 fans, reversible airflow option.
• Confirm transceiver and cable compatibility for your link speeds and reaches.
• Confirm operating temperature range 0°C to 40°C is suitable for your environment.
• Review NVIDIA software release notes for feature availability and supported OS versions.
• Weight: 8.5kg (18.73lb). Use appropriate rack mounting and lifting procedures.

Hong Kong Starsurge Group Co., Limited is a technology-driven provider of network hardware, IT services, and system integration solutions. Founded in 2008, the company serves customers worldwide with products including network switches, NICs, wireless access points, controllers, cables, and related networking equipment.

Backed by an experienced sales and technical team, Starsurge supports industries such as government, healthcare, manufacturing, education, finance, and enterprise. The company also offers IoT solutions, network management systems, custom software development, multilingual support, and global delivery.

With a customer-first approach, Starsurge focuses on reliable quality, responsive service, and tailored solutions that help clients build efficient, scalable, and dependable network infrastructure.