https://evex.one/tags/distribution/ Recent content in Distribution on Eval-EXEC Hugo -- gohugo.io en execvy@gmail.com (EXEC) execvy@gmail.com (EXEC) ©2026 Wed, 22 Apr 2020 10:11:09 +0800 https://evex.one/posts/distribution/6.824/lab_3/ Wed, 22 Apr 2020 10:11:09 +0800 execvy@gmail.com (EXEC) Wed, 22 Apr 2020 10:11:09 +0800 https://evex.one/posts/distribution/6.824/lab_3/ <p>在<code>lab2</code>已经完成了简单的<code>raft</code>, 在<code>lab3</code>里要完成一个<code>key-value</code>数据库。</p> <p>数据库服务端应该有这三个接口：</p> <ul> <li><code>Get(key string) </code></li> <li><code>Put(key, value string)</code></li> <li><code>Append(key, value string)</code> // 如果 key 不存在，那么有等同于 Put</li> </ul> <p>客户端和 <code>Clerk</code>交互， <code>Clerk</code>和<code>raft server</code> 交互。</p> EXEC featured image distribution lab 6.824 https://evex.one/posts/distribution/6.824/lab_2/ Tue, 31 Mar 2020 10:11:09 +0800 execvy@gmail.com (EXEC) Tue, 31 Mar 2020 10:11:09 +0800 https://evex.one/posts/distribution/6.824/lab_2/ <h2 id="raft-简要介绍">raft 简要介绍</h2> <p><code>raft</code>是一种分布式一致性算法，可以先看看这里：<a href="https://raft.github.io/">Raft Consensus Algorithm</a> 的互动小程序来对<code>raft</code>有个感性的认知，然后再读 paper。</p> <h3 id="raft-的角色">raft 的角色</h3> <p>论文：<em>/In Search of an Understandable Consensus Algorithm</em> 里描述的 raft 算法有这几种角色：</p> EXEC featured image distribution lab 6.824 https://evex.one/posts/distribution/6.824/lab_1/ Tue, 24 Mar 2020 23:37:25 +0800 execvy@gmail.com (EXEC) Tue, 24 Mar 2020 23:37:25 +0800 https://evex.one/posts/distribution/6.824/lab_1/ <h2 id="map-reduce-论文阅读">Map-Reduce 论文阅读</h2> <h2 id="分析">分析</h2> <p>实验要求 在 <code>mr/master.go, mr/worker.go, and mr/rpc.go.</code> 中编码，并且只指定了 <code>reduce worker</code>的数量：<code>nreduce</code>。</p> <p><code>master</code>最后要以 <code>go run mrmaster.go pg-*.txt</code> 的方式运行，每一个<code>.txt</code>文件对应一个<code>split</code>，一个<code>split</code>分派一个<code>mapworker</code></p> EXEC featured image distribution lab 6.824