The idea of concurrent processes communicating directly, i.e., without a central service acting as intermediary, has its roots in the development of operating systems in the 1960s. The idea of independent communication led the technical development from early versions of Local Area Networks and finally to the largest distributed system in history, the Internet.
Today, whenever a system is set up, one of the most fundamental decisions to make is concerning its architecture. Especially the way in which its components are organized and related to one another must be chosen thoughtfully. The most popular approaches for a systems architecture are the concepts of centralized systems, distributed systems and their hybrid forms. Since centralized and distributed architectures are antipodes and antipodes always inspired engineers to create hybrid systems, different hybrid architectures were created.
In centralized architectures, all nodes connect to a central node. This central node has the role of the intermediary connecting the surrounding nodes to each other. In contrast, the nodes in a distributed system are connected without any central element of coordination or control.
The advantage of centralized systems is their highly reduced complexity in comparison to a distributed system because of the central node as coordinator of communication and integrity. The central node monitors all system transactions and ensures that each transaction is trust- worthy as a trustee. To ensure this integrity and coordinate the communication between nodes, this central node must provide significant computing power. An additional disad- vantage is that the central node is a single point of failure. Accordingly, when this point fails or is infiltrated, it will stop the whole system from working.
| Figure 2-2: Mixing distributed with centralized architecture (Drescher, 2017, p. 37) | Figure 2-1: Distributed vs. centralized system architecture (Drescher, 2017, p. 27) |
As distributed systems have no central node, they are more reliable in terms of availability and much cheaper because no supercomputer is required as coordinator or trustee. Moreover, distributed systems are independent of central instances and their trustworthiness. Other ad- vantages are the publicity and transparency in distributed systems because of the ability to grow naturally. Each node newly added to the system only needs to be introduced to a single node that is already part of the system. Furthermore, distributed systems can provide higher computing power, since all nodes can combine their computing power to a high-performance network.
One property of distributed systems is both blessing and curse. Because all nodes communicate directly with each other without any coordination or proof of identity there is no chance to identify whether one of the partners in a communication is untrustworthy and spreads malignant or untrue information. This leads to lack of security and integrity. Security and integrity could be achieved in a distributed system by implementing strict access controls where each node that is to join the system must prove its trustworthiness. However, this requires instances for the validation of trustworthiness. These must have the power to decide who is allowed to join the system, again bringing a hidden centrality into the distributed system and therefore directly conflicting with the ideas of publicity, transparency and decentralization. A system like this is a hybrid of centralized and distributed systems and could look like the approach in figure 2-2 on the left hand.
For long time, the problem of implementing security and integrity in fully distributed system remained unsolved, until in 2008 Satoshi Nakamoto published his approach to hand over the regulation of security and integrity to the entirety of nodes in the system as a democracy, which is now commonly known as blockchain.