To be decentralized, or to be distributed: that is the question.

Mark F
5 min readOct 4, 2020
The network graph you never see
The network graph you seldom see

TL;DR

The terms “centralization” + “decentralization” are often describing the governance structure of a single organization. The solution that is often proposed to resolve this is in the form of a DLT network or a Blockchain network, which are distributed databases.

Distributed Ledger Technology (DLT)+ Blockchain platforms are networks that are subsets of distributed databases, so the term “distributed” is typically referring to the architecture of the DLT/Blockchain platform. DLT and Blockchains are not the same thing: Blockchain is a subset of DLT.

The structure of the industry that a DLT/Blockchain platform addresses or creates is a network that could be described with these terms as well.

The Problem:

It goes without saying the the main benefit of platforms like Ethereum is that it provides decentralization… right?

Or… wait… is it distributed?

Our hyper-specialized world doesn’t make it clear, because everyone is talking about the same thing in different contexts and with varying degrees of granularity. Few people clearly define what these mean or what contexts they may be written in, so let’s try to clear things up!

So, What we mean when we talk about a decentralized or distributed network?

We’ll break this down term by term:

What’s a Network?

You’ll have to bear with me for a minute, but decomposing these terms is crucial to using these intelligently when we are discussing new technologies.

When we say a decentralized or distributed network, most of the time we are talking about the concept of Networking in the context of computer systems. For the sake of simplicity, we can use the following definitions:

A system is a collection of subcomponents that are interconnected with one another.

A computer system is a system that interconnects hardware, software, and user data.

A computer network is just a connection that allows multiple computers to communicate with each other.

We can use computer networking to scale a computer system to reach many computers. These networks can be mathematically modelled, optimized, and analyzed in depth. They can be represented by network topologies, which is what the image at the beginning of this article shows.

We overlook another use of the word network. In the context of DLT/Blockchain technologies we may also use the word network when we are representing the structure of the industry that we are addressing.

Here’s where it can get confusing…

Let’s take the example of social media, and say that you want to create a Blockchain based alternative to Twitter that will be resilient to censorship.

A platform like Twitter is a central agent, and is a single entity that can exercise control over its user’s data. Let’s say that you want to create a micro blogging site where no one organization has total control over the platform. In this case, it is the centralization of control that you are addressing. Even though Twitter’s architecture for could technically be “decentralized” or “distributed”, it is the centralization of control that you are addressing.

By the same token, your DLT/Blockchain solution would be a centralized organization that manages the data on a distributed database, but would be decentralized if management decisions were made using a consensus on the network.

As you can see, context is key here. It can get more complicated when you start to consider a new solution to an existing market, which is a network unto itself. These two networks can get tangled and the terms can get confusing if they are new to you.

What kinds of Networks are we dealing with?

I mentioned earlier that networks can be represented by network topologies. This could also be looked at as a graph, which will take you down the dizzying field of graph theory.

https://ocw.mit.edu/courses/civil-and-environmental-engineering/1-022-introduction-to-network-models-fall-2018/lecture-notes/MIT1_022F18_lec2.pdf

It would not be possible to give this the attention it deserves in this article, so we are only going to address the kinds of networks that tend to be compared + discussed in the Blockchain community:

  • Centralized Networks — All nodes are connected under a single authority. Many users going to a single point.
  • Disconnected Decentralized Networks — Many users going to many points, who are not communicating with each other.
  • Connected Decentralized Networks — Many users going to any point in a system, which synchs up with itself.

So when we talk about Ethereum or other platforms offering a decentralized network, we are talking about a connected decentralized network where the control is managed through a consensus within their platform.

But what about the “distributed” side of things?

In most Blockchain/DLT discussions we are talking about how the data is spread across multiple nodes.

So what does “distributed” have to do with this?

To understand where the distributed aspect of Blockchain/DLT comes into play we have to address another layer of confusion, which comes from people incorrectly using the words Blockchain + DLT. These are not one and the same.

Both are types of distributed databases but Blockchain is a subset, or a type of DLT.

  • A distributed database is a database where data is stored across many nodes.
  • DLT is a distributed database with no central administrator.
  • Blockchain is a DLT that uses a specific data structure with specific procedures that it follows. More specifically:

A blockchain is a shared ledger of records organized in ‘blocks’ that are linked together by cryptographic validation

These are implementations of distributed databases, and this is why we use this term.

There is such a thing as distributed computing, which is when a network of computers works together to form one system. One type of distributed computing is the peer-to-peer model, and the Bitcoin whitepaper by Satoshi Nakamoto even defined Bitcoin as a peer-to-peer electronic cash system. The blockchain solution that came in this whitepaper was addressing the shortcomings of the peer-to-peer model. It introduced the consensus protocol, where all nodes must agree to validate that a transaction is valid, and used economic incentives to enforce game theory that drives the nodes to want to be honest. This history between the two terms may also cause some of the confusion.

Ah! This looks familiar:

So we finally find ourselves with Image ( C ), which we have seen on all kinds of Blockchain articles.

Our understanding of this image as a result of reading this article should be that the decentralization pertains to the decision making protocol, while the distributed graph is representing how to data is managed.

Some Caveats

A Blockchain can become centralized if it were to become a private Blockchain. In my opinion, this defeats the “trustless” characteristic of a Blockchain. However, limiting the number of nodes needed to verify in a traditional PoW Blockchain would lower transaction cost. Hyperledger is an example of a private Blockchain.

--

--

Mark F

Decentralized Apps + Markets Enthusiast — I go down the rabbit holes and try to come back with the summaries!