A.3 论文、规范与公开复盘索引
只列本书实际引用或建议延伸阅读的材料。链接以官方/原始来源优先；产品文档会变，引用时注明版本或访问时间。
1 共识与复制
| 材料 | 为什么读 | 链接 |
|---|---|---|
| L. Lamport, The Part-Time Parliament (1998) | Paxos 的原始论文；理解 prepare/accept 的安全性 | https://lamport.azurewebsites.net/pubs/lamport-paxos.pdf |
| L. Lamport, Paxos Made Simple (2001) | 更短的 Paxos 描述 | https://lamport.azurewebsites.net/pubs/paxos-simple.pdf |
| D. Ongaro, J. Ousterhout, In Search of an Understandable Consensus Algorithm (2014) | Raft 原始论文；选举、日志复制、成员变更 | https://raft.github.io/raft.pdf |
| D. Ongaro, Consensus: Bridging Theory and Practice (2014) | Raft 博士论文；实现细节与证明 | https://web.stanford.edu/~ouster/cgi-bin/papers/OngaroPhD.pdf |
| M. Fischer, N. Lynch, M. Paterson, Impossibility of Distributed Consensus with One Faulty Process (1985) | FLP 不可能性；理解为什么需要超时/随机化 | https://dl.acm.org/doi/10.1145/3149.214121 |
| G. DeCandia et al., Dynamo: Amazon’s Highly Available Key-value Store (2007) | 无主复制、quorum、vector clock、 hinted handoff | https://www.allthingsdistributed.com/files/amazon-dynamo-sosp2007.pdf |
| J. C. Corbett et al., Spanner: Google’s Globally-Distributed Database (2012) | TrueTime、外部一致性、2PC 的组合 | https://research.google/pubs/pub39966/ |
2 一致性与验证
| 材料 | 为什么读 | 链接 |
|---|---|---|
| M. Herlihy, J. Wing, Linearizability: A Correctness Condition for Concurrent Objects (1990) | 线性一致的形式定义 | https://dl.acm.org/doi/10.1145/78969.78972 |
| L. Lamport, How to Make a Multiprocessor Computer That Correctly Executes Multiprocess Programs (1979) | 顺序一致的定义 | https://dl.acm.org/doi/10.1109/TC.1979.1675439 |
| P. Bailis et al., Highly Available Transactions: Virtues and Limitations (2014) | 事务与一致性模型的组合边界 | https://www.vldb.org/pvldb/vol7/p181-bailis.pdf |
| S. Gilbert, N. Lynch, Brewer’s Conjecture and the Feasibility of Consistent, Available, Partition-Tolerant Web Services (2002) | CAP 的形式化证明 | https://dl.acm.org/doi/10.1145/564585.564601 |
| D. Abadi, Consistency Tradeoffs in Modern Distributed Database System Design (2012) | PACELC 的提出 | https://www.cs.umd.edu/~abadi/papers/abadi-pacelc.pdf |
| Jepsen, MongoDB 3.4.0-rc3 (2017) | 陈旧读、丢失更新、时钟偏斜的公开分析 | https://jepsen.io/analyses/mongodb-3-4-0-rc3 |
| Jepsen, etcd 3.4.3 (2020) | Raft 实现的严格可串行化、watch 顺序与锁的互斥边界 | https://jepsen.io/analyses/etcd-3.4.3 |
| P. Bailis, K. Kingsbury, The Network Is Reliable (2014) | 真实网络分区比想象中常见 | https://queue.acm.org/detail.cfm?id=2655736 |
3 事务、存储与消息
| 材料 | 为什么读 | 链接 |
|---|---|---|
| H. Berenson et al., A Critique of ANSI SQL Isolation Levels (1995) | 隔离级别与异常的经典辨析 | https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/tr-95-51.pdf |
| M. J. Cahill et al., Serializable Isolation for Snapshot Databases (2008) | PostgreSQL SSI 的理论基础 | https://dl.acm.org/doi/10.1145/1376616.1376690 |
| J. Gray, A. Reuter, Transaction Processing: Concepts and Techniques (1992) | 事务处理的系统教材 | 书店/图书馆 |
| H. Garcia-Molina, K. Salem, Sagas (1987) | Saga 的原始论文与补偿语义 | https://dl.acm.org/doi/10.1145/38713.38742 |
| P. Helland, Life beyond Distributed Transactions (2007) | 大规模系统为什么避免 2PC | https://queue.acm.org/detail.cfm?id=3025012 |
| J. Kreps, N. Narkhede, J. Rao, Kafka: a Distributed Messaging System for Log Processing (2011) | 日志式消息系统的设计 | https://notes.stephenholiday.com/Kafka.pdf |
| Kafka Documentation: Design | ISR、acks、幂等生产者、事务 | https://kafka.apache.org/documentation/#design |
| Amazon S3 Consistency Model | S3 的强一致读后写语义 | https://docs.aws.amazon.com/AmazonS3/latest/userguide/Welcome.html#ConsistencyModel |
4 工程与事故复盘
| 材料 | 为什么读 | 链接 |
|---|---|---|
| D. Kegel, The Tail at Scale (2013) | 长尾、对冲请求、延迟放大 | https://research.google/pubs/pub40801/ |
| B. Beyer et al., Site Reliability Engineering (2016) | SLO、错误预算、on-call | https://sre.google/sre-book/table-of-contents/ |
| LinkedIn Engineering, How LinkedIn customizes Apache Kafka for 7 trillion messages per day (2019) | 分区元数据、rebalance、配额与跨集群复制 | https://engineering.linkedin.com/blog/2019/apache-kafka-trillion-messages |
| AWS, Summary of the Amazon S3 Service Disruption in the Northern Virginia Region (2017) | 重启、缓存重建、重试风暴 | https://aws.amazon.com/message/41926/ |
| GitHub, October 21 Post-Incident Analysis (2018) | 网络分区、MySQL 切换、数据一致性 | https://github.blog/news-insights/company-news/oct21-post-incident-analysis/ |
| AWS, Summary of the AWS Service Event in the US-East-1 Region (2021) | 控制面故障、重试与恢复 | https://aws.amazon.com/message/12721/ |
| GitLab, Postmortem of database outage of January 31, 2017 | 备份、恢复、观测与流程 | https://about.gitlab.com/blog/2017/02/10/postmortem-of-database-outage-of-january-31/ |
| M. Kleppmann, How to do distributed locking (2016) | Redlock 与 fencing 的争论 | https://martin.kleppmann.com/2016/02/08/how-to-do-distributed-locking.html |
| antirez, Is Redlock safe? (2016) | 对上述批评的回应 | http://antirez.com/news/101 |
| Stripe, Idempotent requests | 幂等键的工业实践 | https://docs.stripe.com/api/idempotent_requests |
| U.S. SEC, SEC Charges Knight Capital With Violations of Market Access Rule (2013) | 旧代码路径重复下单、部署不一致 | https://www.sec.gov/news/press-release/2013-222 |
| Cloudflare, Details of the Cloudflare outage on July 2, 2019 | 边缘配置推送与全局故障 | https://blog.cloudflare.com/details-of-the-cloudflare-outage-on-july-2-2019/ |
5 规范
| 规范 | 用途 |
|---|---|
| RFC 9293 (TCP) | 只在解释连接复用/超时时引用；不重写 |
| RFC 9110 (HTTP Semantics) | 方法幂等性、状态码、Retry-After |
| RFC 7231（已被 9110 取代） | 历史引用，理解 Idempotency-Key 的语境 |
| W3C Trace Context | traceparent / tracestate 传播格式 |
| OpenTelemetry Specification | trace/metrics/logs 的 API 与语义约定 |
引用原则
本书不复制论文原文、不搬运事故复盘的长段文字；只提取可核对的事实（时间、组件、失败链条、官方结论）并给出链接。若你发现链接失效，按标题搜索原始来源，不要依赖二手转述。