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
- 实时预览敢让你敲 <script>:自研 Markdown 渲染器的「先转义后拼装」
- How to integrate fingerprinting into your C# application
- 重返C++:从 ref 挖到移动语义,在从 forward 挖到可变参数模板
- 为emlog添加Gravatar头像服务功能
- Android Service:开发自己的通知中心(1):辅助***介绍
- Android Service:开发自己的通知中心(2):辅助***实战
- How to Build an Accordion-Menu using CSS and a bit of jQuery
- Building a table for the nth time - Part 2
- wsf-xxxvi
- Pinned Apps – An App Model for the Web





