How to Deploy 10G Switches in Spine-leaf Architecture?

10G switches or ethernet switches have become ubiquitous in data center applications, acting as both ToR switches and leaf switches in spine-leaf architecture. Influenced by the growing popularity of spine-leaf architectures, the application of 10G leaf switches is very common. Based on this trend, how do you choose a 10G switch to design your data center network fabric? Here is the answer for you.

Why Use 10G Switches for Spine-leaf Network?

10Gb switches or ethernet switches have evolved with significant performance improvements and cost-per-port reduction, having become indispensable devices for network systems. 10Gb switches or ethernet switches serve as leaf switches in spine-leaf architecture, helping to build the network foundation for modern data center applications, while also paving the way for future growth.

When acting as leaf switches, 10G switches can handle L2 and L3 processing, data center bridging, and FCoE for the entire service rack. Also, they can be placed one hop away, eliminating the need to hop up and down the tree design, improving latency, and minimizing bottlenecks. As a result, IT professionals can quickly add new leaf switches to accommodate data center growth without rebuilding the network. When used as a TOR switch, 10G switches have the advantages of lower power consumption, easy expansion, and simplified cabling complexity.

Leaf Switch Features and Requirements

If you are expecting to choose the right leaf switches for your spine-leaf architecture, you first need to have a solid understanding of leaf switches. Leaf switches have two types of ports: uplink ports and downlink ports. Its main purpose is to control traffic between servers and complete traffic forwarding.

In a high-density data center, the uplink bandwidth usually supports 40 or 100G, while the downlink bandwidth is 10G/25G/40G/50G/100G. It is necessary to ensure that the uplink bandwidth is always greater than or equal to the downlink bandwidth. More importantly, the number of leaf switches is relatively large, and more servers are connected. Hence, the virtualization technologies that can be supported by leaf switches should not be underestimated, including VXLAN, stacking, MLAG, IPv4/IPv6, etc. These technologies will make the entire network easy to manage and have higher flexibility to support business expansion.

Concerning the current development trend of network equipment, multi-rate types of equipment continue to emerge on the market. However, 10G leaf switches can support negotiated access of devices with different rates below 10G, which greatly satisfies the connection requirements of multiple rates.

Building Spine-leaf Network with 10G Switches

We first have to know the spine-leaf network architecture: spine switches, leaf switches, and servers. Spine switch ports only connect to leaf switches, which determines the number of leaf switches. The uplink ports on the leaf switches are only used to connect to the spine switches, and the downlink are used to connect servers and other end devices. Together, they determine the size and scalability of the spine-leaf network. The following will take a 40G spine-leaf network as an example to illustrate how to build a spine-leaf architecture with 10GBASE-T switches.

According to the above, a 40G spine-leaf network may need 40G spine switches, leaf switches with 40G QSFP+ uplinks, servers and storage devices. Here we use the S8050-20Q4C with 20 QSFP+ ports and 4 QSFP28 ports as the spine switches and the 10GBASE-T switch S5850-48T4Q with 48-port 10GBASE-T and 4xQSFP+ uplinks as the leaf switches. Servers and storage devices can be either simple Network Attached Storage (NAS) devices or a complex Storage Area Network (SAN) which depends on your actual needs (Read NAS vs SAN: What’s the Difference? for more details).

In the 40G spine-leaf network, links between spine and leaf switches are 40G, and links between leaf switches and servers are 10G. Therefore, the 10GBASE-T switches must connect to the spine switches through the 40G QSFP+ ports by fiber cables, while the copper ports of 10GBASE-T switches connect to servers with Cat6a cables. As mentioned above, the uplink numbers of each leaf switch determine the number of spine switches, and the 40G QSFP+ ports on each spine switch determine the number of leaf switches we can have. In this figure, it can be 2 spine switches and 10G switches. Using 10GBASE-T switch in the spine-leaf network will make the best of your existing Cat6a UTP structured cabling ecosystem. Furthermore, it will help to save costs because copper cable infrastructure is far less expensive than fiber optics.

10G Leaf Switches Recommendation

The 10G switches used as leaf switches not only need to carry the servers' traffic but also need to be responsible for forwarding the traffic to the core layer, so we need high-performance switches to complete the network construction. The FS 10G switches have matured in various aspects, featuring the ability to support high-speed network transmission, as well as rich software and hardware technologies, which help to optimize the flexibility, efficiency, and reliability of the data center network. If you build a spine-leaf network with FS 10G switches, you will improve the overall performance of your network, reduce latency, and achieve zero packet loss.

Conclusion

A spine-leaf architecture brings greater scalability to your data center network and business. No matter what access layer connectivity your needs are, a 10G leaf switch can be your choice, and it can also be used as a TOR switch to help you improve other network or cabling needs. The FS 10G switches have complete operability and security in both software and hardware performance, providing a better network transmission environment for your data center.