Cloud and AI Computing – Clos and Mesh

Apr 10, 2024


Cloud and AI Computing – Clos and Mesh

Apr 10, 2024

By David Wong
Founder of Claruspon


The continue expansion of the Cloud and the AI computing drive the change of the datacenter networks, as it requires a set of features than the Clos topology alone can provide, notably the requirements of low congestion, high throughput and good scalability.

In a series of article, I am about to show you the fundamentals of a leading-edge technology with more desirable networking characteristics than the Clos network alone resulting in more effective network supporting various business goals.

Clos and Mesh

The Clos and Mesh topologies have interesting properties. One of them states that an (n-1, n)-node Clos topology contains1 at least one n-node Mesh topology, where, n-1 refers to the number of fabric switches; n refers to the number of leaf switches, and n is an even number.

Figure 1 – A (3, 4)-node Clos Topology

Today I am going to show you the result of decomposition of an (3, 4)-node Clos topology into a 4-node Mesh topology.

Figure 2-1 – Decomposition 1 of a (3, 4)-node Clos Topology

Figure 2-2 – Decomposition 2 of a (3, 4)-node Clos Topology

Figure 2-3 – Decomposition 3 of a (3, 4)-node Clos Topology

We put back the 3 decompositions and remove all the fabric switches. We have,

Figure 3 – A 4-node Mesh Topology

As once the exclusive paths have been determined, the fabric switches just simply provide fix paths switching and might as well be replaced with simpler point-to-point switching. The point-to-point switching is provided by the leaf switches, and the fabric switches are gone forever from the datacenter network.

The exclusive paths are non-overlapping and do not incur network congestion. The richness of these non-overlapping exclusive paths defines the congestion characteristic of a topology in the modern datacenter.

In the next article "the Roman Market Place", I’ll give an illustration and analysis of the topology versus network congestion.

1 It means that each of the leaf nodes can find an exclusive path to reach one of the other leaf nodes.


David Wong is the founder of Claruspon Systems, Inc., a California company. He is the inventor of 5 US patents in non-minimal path routing, and the Mesh topological networks for datacenter; designer of datacenter switch and modular fabric chassis products. He built a datacenter mini-POD using datacenter switches, and modular fabric chassis as development and validation platform. David is currently building a corporate go-to-market strategy, and venture financing plan. He believes in the can-do spirit, the spirit of innovation, rewards & motivations. David can be contacted at email: [email protected]