status

Understanding Decentralization

Published 26.2.2024

In the realm of blockchain technology, decentralization stands as its cornerstone. Yet, it is often shrouded in a cloud of myths and misconceptions. The concept, while pivotal, is not as straightforward as it seems. Measuring decentralization or comparing different blockchains presents a complex challenge. This article aims to demystify these complexities. Through the use of illustrative images, we will delve into the fundamental principles of decentralization, providing a clearer understanding of this integral feature of blockchain technology. Join us as we embark on this enlightening journey.

Blockchain Is A Distributed And Decentralized Network

The term "decentralized" refers to networks, systems, and applications without a central control point. This concept is key to peer-to-peer (P2P) networks where nodes share data directly, bypassing intermediaries.

In the blockchain industry, the term 'decentralized' is often used loosely. It originally referred only to network architecture, but a blockchain project’s characteristics extend beyond this. They include project management, the team maintaining the source code, governance, and resource distribution for network consensus.

Let's focus first on the network architecture.

A distributed network spreads across a wide area and there are multiple nodes for data processing and storage. This distribution increases efficiency, reliability, availability, integrity, and security.

The image below illustrates a distributed network. Each node, operated by an individual, runs a network client essential for peer-to-peer communication. In the network, each node holds equal power. In the context of decentralization, we could argue that there are 8 entities with equal status in the network.

Nodes collaborate to diffuse and validate transactions. They can perform many other tasks. Every user can interact directly with each other, utilizing a peer-to-peer communication model.

Distributed networks can be decentralized (with no central authority) or centrally controlled.

In the image below you can see the same distributed network as before with 8 nodes. The difference is that all nodes are controlled by a single entity, Alice (red arrows illustrate the control over nodes). A network can have very similar characteristics in terms of efficiency, data availability, and robustness, but it is centralized, which affects data integrity, security, etc.

Alice can decide to shut down all nodes, thereby shutting down the entire network. She can also change the data as she has sole control over it.

A decentralized network lacks a central authority. The heart of decentralization is network consensus, i.e. the rules that nodes use to reach an agreement on changing the state of the ledger at regular intervals. No single entity controls the network, ensuring democratic operation. Anyone is free to join or leave the network.

Instead of a single central authority, the network is decided by a group of entities that share power among themselves. A decentralized network requires protection against Sybil attacks, which we’ll discuss later. Power distribution must be based on the possession of an expensive resource. For PoW networks like Bitcoin, it’s a combination of ASIC hardware and electricity, while for PoS networks like Cardano or Ethereum, it’s ADA and ETH coins respectively.

This creates a system imbalance, as entities typically differ. Those with more wealth, capable of acquiring more of the expensive resources, hold a stronger position in the decentralized network.

The image below illustrates a distributed network, its decentralization determined by the number of PoS coins held by each individual.

It might appear contradictory that entities owning a valuable resource can control a distributed network without operating a node. In many networks, these resource owners can delegate their resources, thus enhancing the power of another node.

Contrary to common belief, decentralization isn’t gauged by the node count in a distributed network. Instead, it’s determined by the distribution and management of an expensive resource among its holders, such as their delegation choices. In a distributed network, certain nodes hold more significance and possess the privilege to generate new blocks.

The term ‘distributed’ refers to node geographical distribution for improved efficiency, availability, and reliability. While anyone should have the ability to contribute to the network with their node, it doesn’t significantly enhance the network’s decentralization. The term ‘decentralized’ focuses on control absence, requiring decision-making power distribution. Decentralization isn’t determined by node ownership, but rather by the possession of an expensive resource.

The Sybil Attack

Blockchain networks are open, allowing anyone to run multiple nodes. However, they’re not inherently immune to Sybil attacks. To resist these, decentralization must be based on owning a scarce/expensive resource, representing a wealth loss risk. This ensures resource owners’ commitment and honesty.

A Sybil attack occurs when an entity creates multiple fake identities to gain control over a network. The attacker aims to influence decisions, manipulate data, or disrupt the network. Decentralization shouldn’t be based solely on node operation ability, as a single entity can cheaply run many nodes.

Consider a network where each connected node has the right to produce a new block and earn a reward. To increase reward chances, one might run multiple clients. Sybil attackers could exploit this, running multiple nodes at a low cost to gain an unfair advantage over honest participants. The image below depicts a scenario where a Sybil attacker operates five nodes. If one entity has majority control over block production or governance, the network is centralized.

Distributed network properties don’t significantly impact decentralization. While nodes protect data and validate transactions, they can’t directly participate in consensus or governance if the operator does not own an expensive resource.

The majority of nodes are just consumers of data, mostly consumers of new blocks. Without owning an expensive resource, they cannot participate in the production of data (blocks).

The production of blocks is governed by those who own the expensive resource. They determine the nodes that will produce new blocks. This design protects against Sybil attacks.

The image shows that only Alice, Bob, Carol, and Dave possess a certain quantity of the valuable resource. Bob has the most, while Dave has the least. Consequently, these four can produce new blocks. Eve, Frank, Grace, and Heidi, lacking this resource, have nodes that merely consume new blocks. Note that a Sybyl attack cannot be performed on this network. The position of the Sybyl attacker would be the same as Eve, i.e. his node would be only a passive consumer of new blocks.

To attack this network, one would need to acquire the expensive resource. This could be done directly by purchasing it, thereby risking their wealth. Alternatively, they could indirectly acquire it through fraudulent means, such as stealing PoS coins or gaining control over a large mining company’s hash rate. However, these indirect methods are typically more complicated.

Control Over A Decentralized Network

In a network, control depends on the owners of the expensive resource, not the node operators. The resource could be owned by a vast number of people, yet only a few nodes in a distributed network are crucial for block production. So, despite a network having thousands of nodes, only a select few actively participate in the network consensus, while the rest are passive data consumers.

If a passive node suddenly becomes unavailable, or even all passive nodes, it will have almost no effect on the network consensus. However, the reverse is not true. If half of the active nodes suddenly become unavailable, the network may not be able to reach network consensus.

From a decentralization perspective, what matters are the resource owners and their decisions on which nodes become vital for the network. Resource owners can use their resources for block production, but this typically requires a substantial amount. More commonly, they delegate the resource to another entity that operates a node actively participating in the network consensus, i.e., producing blocks.

The image shows a decentralized network. Only nodes of Alice, Bob, Carol, and Dave produce blocks. Bob and Carol own a small portion of the expensive resource. All others are active consensus participants. They delegate the resource to one of the operators running a block-producing node (blue arrows show resource delegation). Note that among all delegates, only Olivia, Rupert, and Wendy operate their passive node (which only consumes new blocks).

An expensive resource could be ADA for Cardano or the hash rate for Bitcoin. In Bitcoin’s scenario, you would find an ASIC miner and an electricity source instead of coins. Bitcoin and Cardano work in a fundamentally similar way.

This is what the decentralization of a typical blockchain network looks like. Let's summarize it. 16 entities participate in decentralization, but only 14 of them own the expensive resource. Alice and Dave run block-producing nodes but do not own the expensive resource. There are 7 nodes in the distributed network, but only 4 of them produce blocks. 3 nodes only consume new blocks. In reality, the quantities and proportions among different participants would vary.

Networks often have more delegators than block-producing nodes. Some networks, like Bitcoin, have a vast number of nodes, but only a small fraction actively produce blocks (Bitcoin has roughly only 20 pools, of which 2 are dominant). Conversely, in networks like Cardano, a majority of nodes are involved in the production of blocks. Cardano has thousands of pools. There are staking pool operators that run multiple pools.

Governance

To fully grasp decentralization’s complexity, we must consider the team’s role. Until now, we’ve discussed block production control, which lies with delegators (stakers or miners) and block-producing nodes. Active participants periodically agree on ledger state changes, primarily through new block additions, based on client-defined rules.

It is important to know who controls the rules. This is usually the team that launched the blockchain. That is usually the same team that first published the client.

The figure shows a scenario where one team released three client versions. Operators are free to choose which version of the client to install on their node, thereby determining which protocol rules they will adopt and which will be predominantly enforced throughout the distributed network.

In a decentralized network, the team has only limited control over the rules. All versions of the client (possibly also alternative versions) are publicly available. The team does not (should not) have control over which version of the client the operators install on their nodes.

One of the widespread myths is that when a team has a visible leader, he can change the rules of the network or even shut down the network. However, without owning the necessary amount of resources, it is not possible to change key features of the protocol such as monetary policy.

In the absence of disputes among the team, community, and node operators, there’s a tendency to install the latest client version. This version should maintain key rules unchanged, introducing only minor changes (bug fixes, small improvements) or upgrades that most agree with.

In case of rule disputes, the owners of the valuable resource hold the most influence.

Resource holders delegate their resources to node operators, who are responsible for choosing the version to install. If delegators disagree with the operator’s version choice, they have the option to delegate their resources elsewhere. Therefore, operators must carefully consider their version selection, keeping in mind the preferences of their delegators.

The image illustrates a scenario where two teams, Alpha and Omega, release a client each with distinct and incompatible protocol rules. Three operators who run a block-producing node opt for the Alpha version, while another three choose the Omega version. The resource owners will determine the rules to be enforced. In this particular instance, most of the expensive resources vote for the Omega version.

In the majority of existing networks, one client typically prevails. It’s unusual for several teams to develop an alternate client version, and when they do, this version often represents a minority. This is evident in the case of Ethereum.

Disputes usually arise over different versions produced by the same team. A faction within the team may create a source code fork and modify the rules, making them incompatible with the original version. This new version might be enforced, but it requires backing from the owners of the expensive resource. For the new version to win, the majority of the expensive resources must support it.

Just as with the production of blocks, the network is governed by the owners of the expensive resource. This is crucial in the context of Sybil attacks. If the adoption of the new version was determined by the count of nodes running a certain version of the client, a Sybil attacker, capable of inexpensively assigning a multitude of IP addresses, could potentially dictate the rules of the network. Governance needs protection against Sybil attacks. Honest behavior must be enforced by the potential loss of one's wealth.

Conclusion

Decentralization largely hinges on the distribution of an expensive resource. Only those who delegate can impact the number of viable block-producing nodes (pools) in the network. To probe decentralization, primarily examine the number of individuals possessing a resource, the prevalence of whales among them, and their resource management. In particular, it is important to how many pools the resource is delegated to. Some of the pools may have a dominant position, which is an undesirable phenomenon.

The quantity of passive nodes in the network isn't as crucial for decentralization as the number of block-producing nodes. The distribution of expensive resources is also vital for governance, i.e., upholding rules agreed upon by the majority of the resource holders. The running network is not decided by the team or the CEO, but primarily by the owners of the resource. The availability of multiple alternative clients is beneficial, although not yet the norm.

Featured:

Related articles

Did you enjoy this article? Other great articles by the same author