Decentralized Network

A decentralized network is a network of computers in which control, data and processing are spread across many independent participants, called nodes, rather than concentrated in one central server or organization. Each node can connect to several others, and the network keeps running even if some nodes fail, leave or misbehave. Public blockchains are the best-known decentralized networks in finance, but the idea also underlies file-sharing systems, parts of the internet's routing and some messaging tools.

The term is usually set against a centralized network, in which every participant connects to one hub, such as a bank's core server or a single company's cloud. Engineers sometimes add a third category, the fully distributed network, in which every node is roughly equal; in everyday crypto usage the words "decentralized" and distributed are often used interchangeably. The broader principle is discussed under decentralization.

Picture two ways of passing news around a town. In the first, everyone phones a single switchboard, which then calls everyone else. It is efficient, but if the switchboard breaks or decides to block a message, the news stops. In the second, each person tells a few neighbours, who tell a few more, until the whole town knows. It is messier and slower, but there is no single point where the news can be stopped.

A decentralized network works like the second town. Because many people hold the same information and pass it along, losing one or two of them does not matter much. The cost is that keeping everyone in agreement takes more effort than simply asking the switchboard.

In 1964 the engineer Paul Baran, working at the RAND Corporation, described three network shapes: centralized, with one hub; decentralized, with several hubs each serving a cluster; and distributed, with a mesh of roughly equal links. His work on resilient communications influenced the design of early packet-switched networks. Modern decentralized systems usually rely on peer-to-peer connections, where each node finds and talks to others directly. Messages spread by "gossip", with each node forwarding new information to its peers until it reaches the whole network.

In a blockchain, each node keeps a copy of the ledger and checks new transactions against the rules. A full node verifies everything independently, while lighter clients rely partly on others. Nodes reach agreement on which transactions are valid through a consensus mechanism, designed to work even if some participants are faulty or dishonest, a property known as Byzantine fault tolerance.

• Resilience. With no single point of failure, the network survives outages of individual machines and is harder to knock offline with a distributed denial-of-service attack aimed at one server.
• Censorship resistance. No single operator can easily block users or transactions, a feature described under censorship resistance.
• Open participation. In permissionless networks anyone can run a node, inspect the data and build services on top.
• Shared trust. Participants who do not know each other can rely on a common record without trusting a central administrator.

Decentralized networks are generally slower and more expensive to run than centralized ones, because many machines must store the same data and agree on every change. This tension between decentralization, security and speed is known as the blockchain trilemma. Upgrades are harder, since changes need broad agreement rather than a decision by one owner. Responsibility is also diffuse: when something goes wrong, there may be no company to fix it or compensate users.

Decentralization is a matter of degree. A network may have thousands of nodes but still depend on a few large hosting providers, mining pools or staking services, a pattern discussed under network centralization. Analysts use measures such as the Nakamoto coefficient to estimate how many entities would need to collude to control a network, and a group that gained a majority of mining power or stake could attempt a 51% attack.

The Bitcoin network, launched in 2009, was the first widely used decentralized network for transferring value, with thousands of independent nodes around the world. Ethereum and other smart-contract platforms extended the model to programmable applications. Layered on top of Bitcoin, the Lightning Network is itself a decentralized network of payment channels, in which independent node operators route payments for one another without a central switch. Outside finance, file-sharing systems such as BitTorrent, released in 2001, showed earlier that large peer-to-peer networks could move huge amounts of data without a central server. Outside public blockchains, banks and market infrastructures have tested permissioned networks built on distributed ledger technology, in which a known group of institutions shares control. These are less open but keep the idea that no single member holds the only authoritative copy of the records.