The Three Kinds of Networks (and Why They Matter)

The Three Kinds of Networks (and Why They Matter)

Choosing the wrong network model is disastrous, and avoidable.

The Three Kinds of Networks (and Why They Matter)

There are three different forms of networks in which nodes (whether people, computers, organizations, etc.) can be organized. Each of these forms has strengths and weaknesses, and any strategy that deals with connecting nodes will need to take into account the tradeoffs in the structure of each one.

Disconnected nodes

Nodes

At the outset (and somewhat obviously), nodes that exist without any connection to other nodes are isolated and no network of any form exists.

Centralized

Centralized

The simplest form of a network involves a single, centralized “hub” node to which every other dependent node connects. The classic business model is a centralized network (one seller, many buyers), as are many online content providers (one server, many web browsers).

The advantages of a centralized network include being straightforward to create and ease of standardization and control (the single hub has all the power). An inherent and unavoidable disadvantage to a centralized network is that it creates a single point of failure. If the hub is blocked, destroyed, corrupted, or otherwise rendered inoperable, the entire network goes dark, as all the nodes are connected directly (and exclusively) to the hub.

Decentralized

Decentralized

When more than one centralized network is connected through multiple links, the result is a decentralized network. Examples could include trade associations and computer server farms that provide redundancy and improved reliability. The Internet itself is a complex and decentralized mesh of networks, as there are many servers with hosts connected to them, but the servers are themselves connected to multiple other servers.

Distributed

Distributed

A distributed network is unique in that there are no dedicated hubs. Every node is also a hub, both connecting to other nodes and being connected to by them. Not every node is necessarily connected to every other node, but each node has the ability to connect directly to any other node. This creates an extremely robust, fault-tolerant network. If any node is taken out of the mesh, the impact is minimal because not only are there built-in redundancies in link patterns (providing virtually infinite other routes to other nodes), but it is possible to establish a direct connection between any two nodes in the mesh.

This creates a far more complex network than either decentralized or centralized networks. It is difficult (in some contexts, impossible) to control a distributed network, at least using the traditional “command and control” approaches that are readily available in centralized and decentralized networks. But the absence of a centralized control mechanism does not necessarily result in chaos. Establishing common mission and shared parameters for achieving it enables a high degree of innovation and autonomy at the level of each individual node, while enabling the collective network to accomplish its purpose.

Examples of a distributed network include: ant colonies and Bitcoin. Ant colonies have no control structure: each ant is essentially free to make its own choices about whether to explore new horizons or follow a known trail to food, all for the common good of the colony. Bitcoin implements a completely flat, distributed network, with parameters in the software that enables any node to join the network and, collectively, to accomplish the objective of creating a cryptographically-derived currency.

Conclusion

In an increasingly interconnected world, building and supporting networks of many kinds is essential to success. It is important for strategists and leaders to consider the tradeoffs and implications in network design.