https://plumephp.com/tags/%E5%9F%BA%E7%A1%80/Recent content in 基础 on PlumePHPHugozh-CNFri, 14 Aug 2026 18:01:00 +0800https://plumephp.com/kafka-introduction/Fri, 14 Aug 2026 18:01:00 +0800https://plumephp.com/kafka-introduction/<h2 id="1-核心概念">1. 核心概念</h2> <pre tabindex="0"><code>Producer ──→ Kafka Cluster ──→ Consumer │ ┌────────────┼────────────┐ │ │ │ Broker 1 Broker 2 Broker 3 │ │ │ └── Topic: orders ────────┘ ├── Partition 0 (Leader@B1, Replica@B2) ├── Partition 1 (Leader@B2, Replica@B3) └── Partition 2 (Leader@B3, Replica@B1) </code></pre><table> <thead> <tr> <th>概念</th> <th>说明</th> </tr> </thead> <tbody> <tr> <td><strong>Topic</strong></td> <td>消息分类主题</td> </tr> <tr> <td><strong>Partition</strong></td> <td>Topic 的分片，每个有序不可变日志</td> </tr> <tr> <td><strong>Broker</strong></td> <td>Kafka 服务器节点</td> </tr> <tr> <td><strong>Producer</strong></td> <td>消息生产者</td> </tr> <tr> <td><strong>Consumer</strong></td> <td>消息消费者，通过 Consumer Group 消费</td> </tr> <tr> <td><strong>Offset</strong></td> <td>消息在 Partition 中的位置</td> </tr> <tr> <td><strong>ISR</strong></td> <td>In-Sync Replicas，同步副本集合</td> </tr> </tbody> </table> <h2 id="2-日志存储">2. 日志存储</h2> <pre tabindex="0"><code>Topic: orders / Partition: 0 offset: 0 1 2 3 4 │ │ │ │ │ ┌┴──────┴───────┴───────┴───────┴───────┐ │ msg0 │ msg1 │ msg2 │ msg3 │ msg4 │ ← log segment └───────┴───────┴───────┴───────┴───────┘ 0.log offset 是 Partition 内的局部有序，不是全局有序 </code></pre><h2 id="3-为什么快">3. 为什么快？</h2> <table> <thead> <tr> <th>技术</th> <th>作用</th> </tr> </thead> <tbody> <tr> <td>顺序写磁盘</td> <td>比随机写快 1000x</td> </tr> <tr> <td>零拷贝</td> <td>sendfile，内核直接发数据到 socket</td> </tr> <tr> <td>批量压缩</td> <td>压缩后传输，减少网络 IO</td> </tr> <tr> <td>Page Cache</td> <td>OS 缓存热点数据</td> </tr> </tbody> </table> <h2 id="延伸阅读">延伸阅读</h2> <ul> <li><a href="https://plumephp.com/kafka-producer/">生产者 API 与发送语义</a></li> </ul>