Higher performance, more capabilities: new generation of Eltex data center switches with 400G ports

Since 2020, we have been developing the MES product range for data centers. In 2022, we launched our first devices — MES5400-24 and MES5400-48. In 2023, we introduced MES5500-32 featuring 32 x 100G ports, which became the highest-performance data center switch of the first generation. Later, we built a full product lineup of MES data center switches, offering a variety of port configurations based on 100-gigabit uplinks.

It seems like, 100G per link offers massive performance. However, today’s data center workloads increasingly involve tasks that require even more performance. GPU/TPU clusters for AI training, high-performance computing (HPC systems), distributed storage based on SAN and NVMe-oF, and cloud services (IP Fabric) are already demanding entirely different speeds.

This is confirmed by the dynamics of 400G port adoption, it is significantly higher than it was with 25, 40 or even 100G ports. The 2025 statistics shows that 400G and faster ports already account for a third of the market, and analysts predict the continued growth. This adoption rate highlights the strong demand for high-performance data center infrastructure solutions.

And it’s not just about the performance of an individual link. The number of required connections in a data center is constantly growing. There are too many cables and too many units of equipment involved just to achieve the necessary bandwidth. Furthermore, rack space is being used inefficiently. A standard rack is designed for 42 units, but due to high power consumption of modern hardware, often no more than half the slots are occupied, leaving the rest of the space wasted.

The high-performance ports enable higher connection density: less hardware per rack, simplified cabling, and consequently, easier and more efficient maintenance. 400G switches with breakout mode support enable to solve this issue: a single 400G port can be split into four 100G ports with the total power consumption being lower than that of four separate 100G ports on traditional switches. Lower heat dissipation reduces the load on cooling systems — chillers operate less intensively, resulting in additional energy savings across the entire data center, not just for individual racks.

Eltex already has solutions for the growing demands of modern data centers, and 2026 is the year of new products for IP fabrics. Among the next-generation data center switches with 400G ports, MES5700-32 is in pre-production testing, and MES5600-24 is under development.

Let’s start with MES5600-24. The 1RU form-factor switch features 24 × 100G ports (QSFP28), 8 × 400G ports (QSFP56-DD), and 2 × 10G ports (SFP+). The total bandwidth is 11.2 Tbps. Power is redundant, with two hot-swappable AC/DC power supplies. Cooling is adaptable to the specific needs of the data center: airflow can be configured front-to-back or back-to-front, and the fans are also hot-swappable.

How do we see this model being used?

First, as a powerful Leaf switch. Our primary purpose was to create a device capable of seamlessly connecting the existing periphery to the new high-speed core. If the data center has high-performance servers with 100G uplinks, MES5600-24 collects traffic from them and transmits it to the Spine layer via wide 400G channels.

Secondly, as a Spine switch for mid-sized data centers. 400G ports are for high-speed connections to Super Spine switches or edge equipment, while 100G ports are for connection to Leaf switches, which can be previous-generation Eltex switches, such as the MES5410-48.

If MES5600-24 is about flexibility, MES5700-32 is about pure power and scalability. Here, we've focused entirely on the 400G standard: 32 × 400G QSFP56-DD ports, total bandwidth is 25.6 Tbps.

Two additional 10G SFP+ ports are dedicated to service needs, so as not to take up expensive high-speed ports for management or monitoring. The fault tolerance is identical to MES5600-24: AC/DC power redundancy, front-to-back or back-to-front cooling. Power supplies and fans are hot-swappable.

We primarily see MES5700-32 as a Super Spine switch for collecting and aggregating traffic flows across the entire fabric from downstream Spine and Leaf switches. Additionally, it can act as a Spine switch in fabrics that require connecting high-performance hosts via 100G interfaces.

All MES data center switches support EVPN/VXLAN for building overlay networks over the physical infrastructure. Current IP Fabric capabilities include:

• Anycast Gateway
• Core Isolation Tracking
• DHCP Relay
• EVPN Multihoming
• MAC Mobility
• Symmetric/Asymmetric IRB
• Underlay using IS-IS/OSPF protocols
• VRF Route Leaking
• BUM Traffic Replication (Ingress Replication/Multicast VXLAN)

Another key development area we are working on is support for Data Center Bridging (DCB) protocols. We're implementing them in stages. Currently, the Priority-Based Flow Control (PFC) mechanism has been implemented. Enhanced Transmission Selection (ETS), Explicit Congestion Notification (ECN), Data Center Bridging Exchange (DCBX), and Dynamic Load Balancing (DLB) features are planned in our roadmap and will be implemented in future releases.

In retrospect, we have come a long way in the evolution of MES lineup from basic access switches to terabit monsters. And this path continues. Launching the new generation of MES data center switches is a significant milestone for us. 400G ports in our devices enable to build high-performance networks entirely based on Eltex solutions, ready for current and future workloads.

To learn more, please contact Eltex Sales Department at foreign.sales@eltex-co.ru