Distributed systems

Understanding distributed systems is essential in order to understand blockchain because basically blockchain at its core is a distributed system. More precisely it is a decentralized distributed system.

Distributed systems are a computing paradigm whereby two or more nodes work with each other in a coordinated fashion in order to achieve a common outcome and it's modeled in such a way that end users see it as a single logical platform.

A node can be defined as an individual player in a distributed system. All nodes are capable of sending and receiving messages to and from each other. Nodes can be honest, faulty, or malicious and have their own memory and processor. A node that can exhibit arbitrary behavior is also known as a Byzantine node. This arbitrary behavior can be intentionally malicious, which is detrimental to the operation of the network. Generally, any unexpected behavior of a node on the network can be categorized as Byzantine. This term arbitrarily encompasses any behavior that is unexpected or malicious:

Design of a distributed system; N4 is a Byzantine node, L2 is broken or a slow network link

The main challenge in distributed system design is coordination between nodes and fault tolerance. Even if some of the nodes become faulty or network links break, the distributed system should tolerate this and should continue to work flawlessly in order to achieve the desired result. This has been an area of active research for many years and several algorithms and mechanisms has been proposed to overcome these issues.

Distributed systems are so challenging to design that a theorem known as the CAP theorem has been proved and states that a distributed system cannot have all much desired properties simultaneously. In the next section, a basic introduction to the CAP theorem will be provided.

CAP theorem

This is also known as Brewer's theorem, introduced originally by Eric Brewer as a conjecture in 1998; in 2002 it was proved as a theorem by Seth Gilbert and Nancy Lynch.

The theorem states that any distributed system cannot have Consistency, Availability, and Partition tolerance simultaneously:

• Consistency is a property that ensures that all nodes in a distributed system have a single latest copy of data
• Availability means that the system is up, accessible for use, and is accepting incoming requests and responding with data without any failures as and when required
• Partition tolerance ensures that if a group of nodes fails the distributed system still continues to operate correctly

It has been proven that a distributed system cannot have all the afore mentioned three properties at the same time. This is strange because somehow blockchain manages to achieve all these properties, or does it really? This will be explained later in the chapter where the CAP theorem in the context of blockchain is discussed.

In order to achieve fault tolerance, replication is used. This is a common and widely used method to achieve fault tolerance. Consistency is achieved using consensus algorithms to ensure that all nodes have the same copy of data. This is also called state machine replication. Blockchain is basically a method to achieve state machine replication.

In general there are two types of fault that a node can experience: where a faulty node has simply crashed and where the faulty node can exhibit malicious or inconsistent behavior arbitrarily. This is the type which is difficult to deal with since it can cause confusion due to misleading information.

Byzantine Generals problem

Before discussing consensus in distributed systems, events in history are presented that are precursors to the development of successful and practical consensus mechanisms.

In September 1962, Paul Baran introduced the idea of cryptographic signatures with his paper On distributed communications networks. This is the paper where the concept of decentralized networks was also introduced for the very first time. Then in 1982 a thought experiment was proposed by Lamport et al. whereby a group of army generals who are leading different parts of the Byzantine army are planning to attack or retreat from a city. The only way of communication between them is a messenger and they need to agree to attack at the same time in order to win. The issue is that one or more generals can be traitors and can communicate a misleading message. Therefore there is a need to find a viable mechanism that allows agreement between generals even in the presence of treacherous generals so that the attack can still take place at the same time. As an analogy with distributed systems, generals can be considered as nodes, traitors can be considered Byzantine (malicious) nodes, and the messenger can be thought of as a channel of communication between the generals.

This problem was solved in 1999 by Castro and Liskov who presented the Practical Byzantine Fault Tolerance (PBFT) algorithm. Later on in 2009, the first practical implementation was made with the invention of bitcoin where the Proof of Work (PoW) algorithm was developed as a mechanism to achieve consensus.

Consensus

Consensus is a process of agreement between distrusting nodes on a final state of data. In order to achieve consensus different algorithms can be used. It is easy to reach an agreement between two nodes (for example in client-server systems) but when multiple nodes are participating in a distributed system and they need to agree on a single value it becomes very difficult to achieve consensus. This concept of achieving consensus between multiple nodes is known as distributed consensus.

Consensus mechanisms

A consensus mechanism is a set of steps that are taken by all, or most, nodes in order to agree on a proposed state or value. For more than three decades this concept has been researched by computer scientists in the industry and Academia. Consensus mechanisms have recently come into the limelight and gained much popularity with the advent of bitcoin and blockchain.

There are various requirements which must be met in order to provide the desired results in a consensus mechanism. The following are their requirements with brief descriptions:

• Agreement: All honest nodes decide on the same value.
• Termination: All honest nodes terminate execution of the consensus process and eventually reach a decision.
• Validity: The value agreed upon by all honest nodes must be the same as the initial value proposed by at least one honest node.
• Fault tolerant: The consensus algorithm should be able to run in the presence of faulty or malicious nodes (Byzantine nodes).
• Integrity: This is a requirement where by no node makes the decision more than once. The nodes make decisions only once in a single consensus cycle.

Types of consensus mechanism

There are various types of consensus mechanism; some common types are described as follows:

• Byzantine fault tolerance-based: With no compute intensive operations such as partial hash inversion, this method relies on a simple scheme of nodes that are publishing signed messages. Eventually, when a certain number of messages are received, then an agreement is reached.
• Leader-based consensus mechanisms: This type of mechanism requires nodes to compete for the leader-election lottery and the node that wins it proposes a final value.

Many practical implementations have been proposed such as Paxos, the most famous protocol introduced by Leslie Lamport in 1989. In Paxos nodes are assigned various roles such as Proposer, Acceptor, and Learner. Nodes or processes are named replicas and consensus is achieved in the presence of faulty nodes by agreement among a majority of nodes.

Another alternative to Paxos is RAFT, which works by assigning any of three states, that is, Follower, Candidate, or Leader, to the nodes. A Leader is elected after a candidate node receives enough votes and all changes now have to go through the Leader, who commits the proposed changes once replication on the majority of follower nodes is completed.

More details about the theory of consensus mechanisms from a distributed system point of view is beyond the scope of this chapter. Later in this chapter, a full section is dedicated to the introduction of consensus protocols. Specific algorithms will be discussed in chapters dedicated to bitcoin and other blockchains later in this book.