Network switch topology – Network Glossary Definition

Network switch topology refers to the network design an organization employs, whether in a data center, building or campus. Historically, companies could choose from various simple designs for their enterprise networking solutions, including:

  • Bus topology: Each node is connected to a neighboring node, one after another; also known as a daisy chain.
  • Ring topology: Just as its name implies, the ring topology is similar to bus but forms a complete circuit in a ring format. The most common ring format was Token Ring, where only the node that has a logical token is allowed to send data at any given time.
  • Star topology: Each node in the network connects directly to a central hub, which passes data to other nodes.

By today’s standards, each of these enterprise networking solutions have severely limited function in terms of performance and reliability. As simple network hubs gave way to high speed (Ethernet and Fast Ethernet) devices with Layer 2/Layer 3 switching and routing functionality, it made possible more complex computer network topology designs with dramatic performance and reliability improvements.

A mesh network topology, for example, is when each node can connect via multiple paths to any other node, and every node can participate in the transfer of data.

Eventually, the dominant network switch topology that emerged for client-server and data center networks was a three-tier design, as follows:

  • At the lowest level is the access layer, which consists of switches connected to end user devices, servers, printers and so on.
  • Access layer switches feed into the middle tier switches, known as the aggregate or distribution layer.
  • At the top are core layer switch/routers, which connect access layer switches to each other and to outside services such as the Internet.

Such networks are configured with redundant pathways, mimicking the mesh topology, to ensure there is always more than one network path between any two points in case of a network failure. The networks typically use the Spanning Tree Protocol (STP), which ensures traffic always has two routes to take. However, with STP, all but the primary route is deactivated; only if the primary route fails is the secondary route activated. That presents a limitation that can lead to problems with performance, scalability and fault tolerance, especially under heavy loads.

A newer enterprise network switch topology alternative that overcomes these limitations is the leaf-spine topology.

The leaf-spine topology has just two layers: leaf and spine switches. It was originally implemented in data centers, because it’s a good fit for the “east-west” nature of data center traffic. That is, traffic flows largely between servers and storage systems that are located next to each other from a logical perspective, as opposed to “north-south” – meaning outside to other locations or downstream to client devices.

In the leaf-spine topology, the leaf switches connect to end devices such as servers, firewalls and edge routers. Each leaf switch connects to multiple spine switches – perhaps to every spine switch in a data center network.

With the leaf-spine topology, all switches are only a single hop from one another, bringing significant performance improvements as compared to three-tier networks that use STP.

While it’s clear that the leaf-spine switch topology offers benefits for data center networks, some vendors have technology that extends those benefits to the wider enterprise network.

The white box networking vendor Pica8, for example, extends the leaf-spine topology throughout the enterprise with its PicaPilot technology. When used with the Pica8 PICOS network operating system, PicaPilot enables hundreds or thousands of white box network switches to be managed as one single, logical switch, with a single IP address. The switches retain all their Layer 2/Layer 3 switching and routing features, while also supporting software-defined network (SDN) protocols – all on the same ports. It’s a highly scalable, fault-tolerant architecture that ensures enterprise-grade performance and reliability.

PICOS is also an open, standards-based NOS, meaning it can be ported to a number of compatible white box switches. That gives the enterprise a great deal of choice in the hardware that best meets their price and performance requirements.