BFT in a Distributed Compute and/or Storage System - A Simple Introduction

Distributed Systems - physical

Distributed systems is breaking up a large computer/storage (the slash means and/or) into smaller ones. The smaller compute/storage is called a node. Distributed systems can make an entire larger system safer (no single point of failure), more powerful (break up a big compute task into smaller ones), scalable.  Apache Hadoop, where storage is spread across a wide network of storage nodes, is an example of a distributed (storage and processing) system.

But these nodes need to be in sync (aka consistent). And if the nodeA and nodeB have differences in opinion on what the final state of the system is, who is right?  In a space capsule example - two computers are computing the exact time to fire thrusters. ComputerA says 10:01, ComputerB says 10:04. Who is right?

This is why distributed systems have a consensus, or voting, mechanism.  And BFT is the science behind making voting mechanism. BFT is a voting mechanism that allows for multiple nodes to reach a consensus.  BFT can help to make a distributed system 1) failure tolerant - so that if a node on the distributed system fails, the other working nodes can take over quickly 2) malice tolerant - so that if a node on the distributed system is injecting wrong information, it will be voted out by other good nodes.

Decentralization - control 

Decentralization is a CONTROL concept, not a physical concept. So distributed is a physical concept. In a network that is responsible for control, decisions needs to be made and agreed upon. What is the final state of Joe's bank account? Should the reverse thrusters be fired at 10:01 or 10:03? BFT is critical in a decentralized network - to ensure that far away nodes can make decisions.  BFT needs to handle both failure and malice. Malice can come in many forms : from sending the wrong data (Joe owes me $1,000 instead of $1) to jamming the network. Examples of the latter include email DDoS - remedied by asking the sending to do some work first before the sender's email is read by the receiver. 

BFT in Blockchain

Blockchain uses several different ways to keep its distributed nodes in sync (consistent). BFT is one. Another way is to use Proof of Work (Nakamoto Consensus) used by Bitcoin. Rather than using BFT, Bitcoin wants to see proof that the node did work before its vote is accepted. The work that a node needs to perform is very compute intensive, using up much energy and is frowned upon as a way to solve a consensus. Proof of Stake is slated to replace PoW.

Blockchain Refresh

- append only, making it immutable

- time stamped

- consensus 

- hashing : compress transactions, SHA256

- digital signature : private key (held secretly in a wallet) and public key (forms public address)

- chain : a block of data contains multiple transaction hash; and the following block has a hash of the current block so that changes to the current block will impact all blocks 

- store efficiently : in Merkle Tree (hash)

Conclusion

BFT is the science and art of making nodes in a decentralized environment stay sync (consistent) by providing a consensus (voting) mechanism that works robustly (failures are handled) and safely (malice proof). BFT is used in blockchain.