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.
Current
·
C2
·2026-11
Deterministic Simulation Testing
Find Your Interest
- 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
- Android 无障碍服务导致的整机卡顿案例分析
- 重返C++:从 ref 挖到移动语义,在从 forward 挖到可变参数模板
- Integrating in-app-reviews with Trusted Web Activity
- Testing Loading Skeletons
- Never Use The Page As The Source Of Truth
- Android 应用启动优化 - 一种 DelayLoad 的实现和原理(上篇)
- Android accessibility: roles and TalkBack
- 重返C++:C++ 类型擦除
Current
·
C2
·2026-11





