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
Event trace
Timeline
Time Slices
- T0 3 篇 2026-11-17 22:33 Why yes, I do have a patent on a time machine In a deterministic simulation, you can debug with time travel Deterministic Simulation Testing
- T+2h 3 篇 2026-11-18 00:33 Why yes, I do have a patent on a time machine In a deterministic simulation, you can debug with time travel Deterministic Simulation Testing
- T+12h 3 篇 2026-11-18 10:33 Why yes, I do have a patent on a time machine In a deterministic simulation, you can debug with time travel Deterministic Simulation Testing
- T+24h 3 篇 2026-11-18 22:33 Why yes, I do have a patent on a time machine In a deterministic simulation, you can debug with time travel Deterministic Simulation Testing
Conclusion
该事件汇聚 3 个来源,平均可信度 37%
Confidence: medium
Find Your Interest
Current
·
C2
·2026-11





