Explore the intersection of Aeron and Raft in this 44-minute conference talk from GOTO Chicago 2018. Dive into consensus protocols in distributed systems, understanding how they enable a common view of shared state. Learn about Raft's design for understandability and its success in becoming widely adopted. Discover the functional theory behind consensus, with state treated as immutable. Investigate why Aeron was chosen, exploring its guaranteed delivery, architect demand, and need-to-know approach. Examine performance considerations, including RPC, concurrency, and parallelism. Gain insights into pipelining, multicast, monotonic functions, and efficient timer scheduling. Explore real-world applications through examples of distributed file systems and session management. Conclude with a look at the implementation roadmap for integrating these powerful technologies.
Overview
Syllabus
Intro
What does consensus mean
Secondary definition of consensus
What is consensus
Raft
What is Raft
How the election works
Safety guarantees
Functional theory
State as immutable
Why Aeron
Guaranteed Delivery
Architect Demand
NeedtoKnow Approach
Guaranteed Processing
Single Service
Open Source
Do servers crash
Weird Java code
Performance
RPC
concurrency and parallelism
hard work
Speculation engines
Pipelining
Consensus
Pipeline
Example
bytebuffers
bug
message index
byte index
Multicast
Monotonic functions
Efficiency
Timers
Scheduling Timers
Scheduling Timer API
Timer Wheel
Complexities
Real Code
Session Status
Continued Action
Distributed File System
Processing
Implementation
Roadmap
Taught by
GOTO Conferences
