Explore the robustness of Solid State Drives (SSDs) under power fault conditions in this 32-minute conference talk from FAST '13. Delve into a new methodology for exposing reliability issues in block devices during power failures, a common occurrence in data centers. Learn about the specially-designed hardware for injecting power faults, workloads for stressing storage components, and techniques for detecting various types of failures. Discover the surprising results from testing fifteen commodity SSDs from five different vendors, revealing issues such as bit corruption, shorn writes, unserializable writes, metadata corruption, and total device failure. Gain insights into the challenges posed by modern storage technology and the importance of understanding component behavior during power faults for designing robust storage systems.
Understanding the Robustness of SSDs under Power Fault
Overview
Syllabus
Intro
Power Faults a threat never gone
Simple Failures
Complex Failure: Unserializable Writes
Design
What to Write?
Special Record Format - allows detecting all types of failures
Advanced FTL: Compression
Randomization of Record Content - avoid interference of compression
Deriving Completion-time Partial Order - a key step of unserializable writes detection
Power Fault Injection
Experimental Environment
Summary of Observations
Shorn Writes: Subpage Programming
Serialization Errors: Avg. Numbers Per Fault
Serialization Errors: Patterns Over Time
Metadata Corruption
Conclusion
Taught by
USENIX
