Distributed systems are hard to test because partial failures overlap with live traffic and the same fault can produce different execution traces on different runs. Unmesh Joshi describes how deterministic simulation testing addresses this by replacing real clocks, networks, and storage with simulated components driven by a single-threaded execution loop. This allows simulating both hard failures and routine slowdowns—such as network and disk latency spikes—and more importantly makes every execution reproducible. This article explains the technique using a quorum key-value store as a running example, with code from Tickloom, a Java implementation of the approach.
Deterministic Simulation Testing
Find Your Interest
- 🌳 Linux - tree 生成目录树「Note」
- TopDocs:一款美观实用的在线文档编辑系统,支持Markdown语法
- How to Build an Accordion-Menu using CSS and a bit of jQuery
- Scalable Continuous Integration for iOS
- 实时预览敢让你敲 <script>:自研 Markdown 渲染器的「先转义后拼装」
- Write CSS. Not too much. Mostly scoped.
- Building a table for the nth time - Part 2
- Advanced CSS-Only HTML Form Styling
- SVG Partial Blur and iOS Style Translucency
- Pinned Apps – An App Model for the Web





