Arista Networks launches two new platforms compatible with 400 Gigabit Ethernet

The new 400G switches deliver the performance that large-scale cloud networks and data centers need for the growth of high-quality video, artificial intelligence or serverless computing applications.

Arista Networks has once again overcome a new barrier by launching two new platforms compatible with 400 Gigabit Ethernet. The new 400G switches offer the performance that large scale cloud networks and the data centers they need for the growth of high-quality video, artificial intelligence or serverless computing applications.

The series Arista 7060X4 uses industry standard optics for 400G and offers a choice of two optical module formats: OSFP y QSFP-DD. Both form factors offer a wide range of 400G optics and cables, including optics that provide compatibility with 100G ports. This offers customers flexibility in building high-density leaf-spine networks to support the most demanding emerging applications, including 100G server connectivity and 400G leaf-spine connectivity.

Andreas Bechtolsheim, president and chief development officer of Arista, highlights that these new 400G switching platforms "provide four times the performance, double the price/performance ratio and double the energy efficiency of our previous 100G platforms. These products will respond to the growing demand for more bandwidth in cloud networks, allowing cloud service providers to build much more scalable networks."

The new platforms, based on the Broadcom 12.8 Tbps Tomahawk 3 Silicon model, support 32 400G ports in a compact 1U chassis.

In addition, each 400G port can be divided into four 100G ports, supporting a total of 128 100G ports in a 1U chassis. Arista EOS traffic management and load balancing improvements take advantage of on-chip improvements in buffering and routing resources.

Migrating network tiers to 400G with 100G high-density computing maximizes the performance of next-generation applications, such as artificial intelligence, machine learning, and serverless computing, by increasing network speeds and eliminating oversubscription between leaf and spine tiers.

Main features

• 4x more bandwidth density compared to existing 1U 100G switches.
• Shared buffer optimized to reduce application latency for AI and storage workloads.
• Elephant flow detector for advanced traffic management
• Dynamic Load Balancing and Dynamic Group Multipath to improve performance in large cloud and HPC networks.
• Increased routing capacity combined with 2X increased IO for hyperscale backbone
• Optimized low power with uncompromising performance at 8 Tbps
• Consistent EOS support including cloud-based automation, analytics and availability
• Option of OSFP and QSFP-DD optics supporting 50G/100G/200G and 400G speeds