Decentralization

What is a Decentralized System?

Decentralized systems are the opposite of centralized systems. The ruling power concentrated in a centralized system is split and distributed to a large number, if not all, participants of a system. For each decision that must be made, all participants must be included. They have to come to a consensus that holds true for all members. All participants are considered equals and must abide by the system’s rules.

The consensus models can vary between systems. Simple majority or 2/3 majority votes are relatively simple examples. More complex schemes with various layers can be applied if a given environment or situation demands it.

Democratic systems are generally viewed as decentralized. A good example is voting among your friends about which movie to watch at the cinema. Everybody in your group is involved in the decision-making process, and everyone knows and accepts the protocol; raise your hand, and the majority vote wins. If the cool kid in your group is sick that day, the system still works, and the rest of you can enjoy a movie without being forced to follow the preferences of a single ruler. The group also works if multiple people do not show up or your friends bring some more friends along. If someone categorically does not want to watch that horror movie, you could also switch the consensus model to vote for a film acceptable to all and thus reduce friction within the group.

Modern western democracies are also considered decentralized, as leaders, e.g., the president, the parliament, and others, are elected by all eligible citizens. However, the citizens of a given country are not directly included in each decision the government makes. They must trust their representatives to honor their campaign pledges and vote in their interest in parliament. But from all the political dramas on TV, fiction or not, we know this trust is often misplaced.

Thus, democracies are relatively centralized, as the elected officials are in power and can use this power in a centralized way to either make good on their promises or solely act in their own interests. The separation of powers, oversight, and federated systems can mitigate these risks but cannot eliminate them entirely.

Consequently, there are various levels of decentralization ranging from rather centralized to very decentralized. For example, a system in which each and every decision is put to the vote for everyone to participate holds a higher degree of decentralization than electing a ruler who will make the ‘right’ choices.

However, holding a vote on practically everything does not scale. If the number of participants or decisions grows substantially, it becomes extremely cumbersome to conduct polls. This holds especially true in the real world, where every vote takes quite some time: people need to understand their choices, sending messages takes time, counting the ballots, campaigning, etc. Centralization, on the other, hand reduces these effects at a cost. By centralizing to a certain degree, every participant loses part of their influence as this power is given to some centralized entity. Nevertheless, this entity can make and execute decisions faster as the process is more streamlined.

When setting up a new system, one must find an appropriate balance between scalability and security:

• Who should be in control?
• Is centralization a problem?
• What are the consequences if someone abuses their powers?
• How many, and how often does the system have to make decisions?

Decentralized Software Systems

Decentralized software systems have been around for a very long time. Peer-to-peer file-sharing systems are the most notable instances. For example, BitTorrent became very popular in the mid-2000s and remains one of the top file-sharing protocols.

Each node in the network can share files at their free will and ask their peers for the files they are looking for. Files are identified via content hashing, making the system resistant to manipulation. While downloading a file, the downloader can already start sharing completed chunks of the given file. This allows the system to scale reasonably well if enough nodes participate in distributing data. Lists of peers and content are exchanged and propagated throughout the network. There is no global, shared state of the network as it is in constant flow due to clients leaving and joining the system and contributing different sets of data.

While file transfers are conducted between peers directly, the system is somewhat centralized when using centralized trackers. These trackers are used for the initial seeding of files and for bootstrapping of peers, leaving the system with a point of weakness.

Another famous example is blockchains. They differ from p2p file sharing networks as they must maintain a global state all participants agree on. To achieve this, they leverage consensus models like Proof of Work or Proof of Stake to eventually find the ‘correct’ state. Once agreed on, this state holds true forever as later changes are only appended to the blockchain. This fact makes blockchains the perfect system to store valid information indefinitely. Furthermore, the agreed state is shared throughout the network. This creates severalfold redundancy, allowing the blockchain to stay available even if numerous nodes fails.

However, as previously stated, finding a consensus among a large set of participants scales poorly. Moreover, since Blockchains are generally bound by the Blockchain Trilemma, they must sacrifice scalability, security, or decentralization in favor of the other two.

Protocols and system rules are defined in a decentralized system, which participants must adhere to. Within these given parameters, users can use their own protocol implementations. This allows them to customize the system to their needs and preferences. For example, they could use an implementation in another programming language that performs better in a given environment. In addition, they could choose not to interact with or favor certain peers over others. Such ‘optimizations’ might improve their use case for participating in a system, like getting better transfer speeds on BitTorrent or not sharing ‘illegal’ content publicly.

Risks

A decentralized system cannot be controlled by a single party or node. This fact creates two system risks: hostile takeovers and potential system misuse.

If some party gains control over a large enough number of participants, they can block decisions and bring the system to a halt depending on the consensus model. Furthermore, a malicious entity with a controlling majority can control the entire system and effectively make decisions independently, disregarding any other participants. This might allow them to create invalid states by accepting and distributing false computations and data. The remaining independent nodes would reject this corrupt data as it violates the system rules. However, the ruling majority propagating incorrect information wins and effectively breaks the system.

Thus, a decentralized system must try to avoid the concentration of power among nodes by, for example, tweaking the consensus model or the incentive system for participation. Discussions on Proof of Work (PoW) vs. Proof of Stake (PoS) within the crypto ecosystem address this issue.

PoW networks like Bitcoin are secured by computing power. The more computing power someone holds, the more influence they have on the system. As numerous Bitcoin miners who contribute their computing power, a potential attacker must invest heavily in specialized computer hardware to have a significant impact. At the same time, anyone can start contributing their computer power to the network without permission or any other significant obstacles. And with growing interest in the system, more entities are willing to join the network not solely to earn from mining but to secure the system. This means that hurdle to control the majority of computing power in the network becomes ever higher, essentially an arms race with millions of competitors and no end.

PoS systems like Ethereum, as of 2022, are controlled by nodes that hold Ether. The more Ether some entity owns, the more likely they will be selected to write new blocks to the blockchain.¹ As Ether can be bought on the open market and with fiat currency² on CEXes³, malicious entities with enough capital can purchase a controlling majority. This is especially dangerous as entities like banks or governments probably have an interest in controlling systems like Ethereum.⁴

BitTorrent is more or less resistant to such attacks as there is no such thing as a controlling majority due to the absence of a global state. The network is known for the distribution of pirated content. Technically there is no way to prevent the sharing of such content on the protocol level. However, it is up to each node in the network to decide which content they want to share. This might reduce the scaling abilities for that given content, but it cannot be stopped as long as there are participants left wishing to share that content.

Likewise, data written to a public permissionless blockchain can never be removed. There are no administrators. Nobody holds such powers within the system. Even court orders or other measures that would work in the real world cannot be applied without fatally disrupting the system.

But due to the open nature of such decentralized systems, real-world entities like law enforcement can probably find the people who run the nodes in a network to apply pressure.

Benefits

The core benefit of a decentralized system run and owned by the community lies in the capability to resist manipulation. It strictly follows the rules defined in its protocol and cannot break these or make exceptions if they result in a ‘better’ outcome. Of course, the community as a whole might decide to make corresponding changes, but no sole participant can make these alterations. If only a minority adopts the modifications, the non-adopting majority will disregard conflicting decisions. The system will continue to work in its previously established form. Changes that only benefit a minority will not be adopted.

Due to its distributed nature, it can continue to work and serve its purpose even if large parts of the system fail. This might be caused by technical problems like network failures or attacks on individual nodes. There is no single point of failure that can bring the system to a halt. However, degraded service or performance might be possible.

Consequently, a truly decentralized system cannot be shut down as long as nodes want to continue participating in the network, at least from a technical point of view. The effectiveness and efficiency of a system with meager participation might come into question leaving the system effectively for dead. This emphasizes that such a network lives and flourishes with participation.