raft

[ytgui]

https://raft.github.io/ https://raft.github.io/raft.pdf

[ytgui]

Raft

• a consensus algorithm, designed to be easy to understand
• equivalent to Paxos in fault-tolerance and performance
• different to Paxos, Raft has independent subproblems, production ready.

Consensus

• multiple servers agreeing on values in fault-tolerant distributed systems
• once they reach a decision on a value, that decision is final
• consensus algorithms make progress when any majority of their servers is available, and stop making progress if nore servers fail
• never return an incorrect result

Consensus typically arises in the context of replicated state machines (多状态机场景), each server has a state machine and a log.
The state machine is the component that we want to make fault-tolerant, such as a hash table.
In client side, server is a single realiable state machine.
A consensus algorithm is used to agree on the commands in the servers' logs.

[ytgui]

Raft

Introduction

• Strong Leader: log entries only flow from the leader to other servers.
• Leader election: randomized timers to elect leader, resolving conflicts simply and rapidly.
• Membership changes: new joint consensus approach allow cluster to continue operating normally during configuration changes.

Replicated state machines problem