Explore the Saga Pattern for distributed systems in this 34-minute conference talk from GOTO Chicago 2015. Dive into alternatives to traditional ACID transactions and two-phase commit protocols for complex, collaborative processing. Learn about stateless systems, consistency models, and failure conditions before delving into the origins and implementation of Sagas. Discover compensating transactions, atomicity, and failure management techniques. Examine practical examples of successful and failed Sagas, both in single database and distributed environments. Investigate distributed log systems, execution coordinators, and messaging semantics. Gain insights from real-world applications, including Halo game statistics. Evaluate tradeoffs and recap key concepts to enhance your understanding of distributed system design and implementation.
Overview
Syllabus
Introduction
Stateless Systems
Consistency
Twophase commit
Prepare phase
Commit phase
Failure conditions
spanner
can we do better
where Sagas come from
what is a Saga
compensating transactions
successful
atomicity
failure management
single database
Saga execution coordinator
Saw log
Successful Saga Example
Failed Sagas
Successful Sagas
Distributed Systems
Distributed Saga
Distributed Log
Execution Coordinator
Starting a Saga
When to Compensate
Request Messaging Semantics
Application Semantics
Distributed Sagas
Master Chief
Halo Statistics
Halo Actors
Forward Recovery
Game Drain
Recovery
Tradeoffs
Recap
Taught by
GOTO Conferences
