分布式系统

A distributed system is one in which the failure of a computer you didn't even know existed can render your own computer unusable.

---

在一个系统中有一台你从未见过的机器出了问题, 却能致使你的机器也无法使用, 那么它就是一个分布式系统 ------ 莱斯利.兰伯特, 1987

---

为什么我们需要分布式系统？

• tolerate node failure 容忍节点故障

• scalability (processing more requests than a single machine) 扩展性

• latency (placing replicas geographically closer to users) 延迟

---

分布式系统的麻烦

• Unreliable Networks 不可靠的网络

  网络可能会丢包，乱序，超时

• Unreliable Clocks 不可靠的时钟

  不同机器的时钟不一致，另外时钟会出现偏移

---

Models describe the key properties of a distributed system in a precise manner.

• System model (synchronous, partially synchronous, asynchronous)

• Failure model (crash-fail, crash-recovery, Byzantine)

• Consistency model (strong, weak, eventual)

---

Theorem 定理 (maybe?)

• FLP

  FLP proved that consensus cannot be achieved in asynchronous distributed systems if failures are possible.

• CAP

  Available or Consistent when Partitioned

---

分布式系统一个首要问题是进程间通信

• RPC (Thrift, grpc, Rest/Http, Dubbo…)

• MQ (AMQP based: RabbitMQ; log based: Kafka)

---

其次，进程间需要同步：互斥（分布式锁）和分布式事务（原子提交）

• Distributed Lock

• Distributed Transaction

---

共识算法 Consensus Algorithm

Getting a group of nodes to agree on a given value is described as the consensus problem

共识问题: 一组节点就选择一个 value 达成共识 (single decree).

• Paxos

• Raft

• Zookeeper

• Etcd

---

分而治之

• Replication 副本

• Partitioning 分区

进一步阅读

Online Blog & Articles

• Distributed Systems for fun and profit, mixu

  概述分布式系统的多个主题

• Fallacies of Distributed Computing Explained

  L Peter Deutsch 在 Sun 公司发表的分布式计算的 7 个谬论

---

Books

• Martin Kleppmann: Designing Data Intensive Applications - The Big Ideas Behind Reliable, Scalable And Maintainable System

  设计数据密集型应用, 对于实践有非常好的指导价值, 建议作为入门读物

• Joe Armstrong: Making Reliable Distributed Systems in the presence of software errors

  中译本: 面对软件错误构建可靠的分布式系统

• Architect for Scale - High Availability For Your Growing Applications

---

• Andrew S. Tanebaum: Distributed Systems Principles And Paradigms

  中译本, 分布式系统原理与范型; 偏理论

• Distributed Operating Systems 分布式操作系统

• Introduction to Reliable and Secure Distributed Programming

• Nancy A. Lynch: Distributed Algorithms

---

How To Architect

• Kate Matsudaira: Scalable Web Architecture and Distributed Systems

  This article seeks to cover some of the key issues to consider when designing large websites, as well as some of the building blocks used to achieve these goals.

• Google: Designs, Lessons and Advice from Building Large Distributed Systems

• Dan Pritchett: Architecting for Latency, eBay, October 2007

• Architecting for scale, Beautiful Architecture, Chapter 3

---

• High Availability Concepts and Best Practices in Oracle