Distributed System Theory for Software Engineers

Failure Modes

• [x] FAILURE MODES IN DISTRIBUTED SYSTEMS

Clock

How you decide whether an event happened before another event in the absence of any shared clock

• [x] Lamport clocks
• [x] Vector clock
• [x] 分布式数据库中为什么要使用 Vector Clock？

Fault Tolerance

• The quorum technique for ensuring single-copy serializability
  • [ ]Quorum wikipedia
  • [ ]A Quorum-based Commit Protocol
• different fault-tolerance properties
  • [x] Two-Phase Commit
  • [x] Three-phase Commit
  • [ ] Increasing the Resilience of Distributed and Replicated Database Systems
  • [ ] Analysis and Verification of Two-Phase Commit & Three-Phase Commit Protocols
  • [ ] Sinfonia: a new paradigm for building scalable distributed systems
  • [ ] Paxos
• How eventual consistency, and other techniques, seek to avoid this tension at the cost of weaker guarantees about system behaviour
  • [ ] Dynamo
  • [ ] Life Beyond Transactions

Basic primitives

• Leader election
  • [x] Bully algorithm
• Consistent snapshotting
  • [x] Distributed Snapshots: Determining Global States of Distributed Systems
• Consensus
  • 2PC
  • Paxos
• Distributed state machine replication
  • State machine replication wikipedia
  • Lampson’s paper: How to Build a Highly Available System Using Consensus

Fundamental Results

• CAP Theorem
• Papers We Love SF talk

Real systems

Google

• GFS
• Spanner
• F1
• Chubby
• BigTable
• MillWhell
• Omeg
• Dapper
• Paxos Made Live
• The Tail At Scale

Not Google

• Dryad
• Cassandra
• Ceph
• RAMCloud
• HyperDex
• PNUTS

First Step

• Distributed systems for fun and profit by Mikito Takada
• Notes on Distributed Systems for Young Bloods
• A Note on Distributed Computing
• Fallacies of distributed computing

Reference

• machine-learning-for-software-engineers
• Distributed systems theory for the distributed systems engineer