In my next post I will be discussing the Byzantine General's Problem (BGP), a classic computer science problem in creating distributed, decentralized peer-to-peer networks. If you're new to cryptocurrency and blockchain technology it helps to build your level of understanding from the ground up. Bitcoin, Neo, and Ethereum are popular network protocols attempting to solve BGP with a decentralized, distributed network. If you've been following my posts you will see I post a variety of topics from ICO reviews (primarily to pique interest to my site), reviews of promising projects already listed on an exchange (see my posts on Qtum and Ontology), and a series of op-ed articles around the trending topics in cryptocurrencies. With that said, let's get into system design.
Designing The System Network
When designing a network, designers are faced with three options: distributed, centralized, and decentralized networks. Depending on the goals of the organization, one network design is optimal over another. Below is an illustration of these different network architctures.
Let's define each of these systems:
Distributed systems have no single authority. Each of the dots in the diagram is a node, or a network participant. Each node is connected directly to every other node. In non-computer network terms, we can imagine a vote where each voter has equal power to one another.
Centralized systems has a single authority. All nodes obey the commands of the dot colored yellow in the diagram. In non-computer network terms, we can imagine a King or Emperor commanding their nation.
Decentralized systems, like distributed systems, have no single authority. The difference here is there are layers of nodes. The blue nodes, or end nodes, connect to secondary nodes (marked in yellow). The secondary nodes connect to one another. It should be noted there is a high degree of variability in how these systems are constructed but my commentary stays faithful to the image above.
So which architecture should designers choose
Bootstrapping, Creating, Updating a Network
[Highest] Centralized systems can be created quickly. They can also be updated quickly.
[Medium] Once you leave Centralized systems, decisions become slower as decisions and actions start to slow down in Decentralized systems.
[Low] Distributed systems are by far the slowest.
Maintenance
[Easy] Centralized systems are easier to maintain since there is only one point of failure.
[Medium] Decentralized have more points of failure but fewer than distributed systems
[Hard] Distributed systems are harder to maintain due to a large number of nodes.
Stability / Fault Tolerance
[Highest Risk] Centralized systems are vulnerable to a failing leader. In computer networks, if the central server is down the entire network goes with it.
[Medium Risk] Decentralized systems are less vulnerable. One secondary node goes down but the network goes on. There are many other secondary nodes to keep fighting the good fight.
[Low Risk] Distributed networks are the least vulnerable. Kill one head, and two more grow in its place. Hail Hydra!
Evolving the Network
[Slowest] Centralized systems evolve the slowest since they have one single authority to develop new ideas and incorporate it into the network.
[Medium] Decentralized systems have more points to bring in new data to evolve the network.
[Fastest] Distributed has even more access points than decentralized systems.
Scalability
[Low] Centralized systems are low scalability options.
[Medium] Decentralized has high scalability with finite capacity.
[High] Distributed systems have infinite scaling limits.
Now that you've read a quick primer on system design, I will explore later this week how these systems attempt to solve the Byzantine General's Problem.
