Research project on detecting and correcting transient hardware errors (bit flips) during query processing in main-memory-centric database systems (especially, column stores).
AHEAD is our in-memory column store prototype which incorporates AN Coding for detecting (transient) hardware errors during query processing – at run-time! The respective paper “AHEAD: Adaptable Data Hardening for On-the-Fly Hardware Error Detection during Database Query Processing” was accepted at SIGMOD’18.
We will gradually fill these sections with more contents.
When do we do bit flip detection?
We evaluated AHEAD using the Star Schema Benchmark (SSB)[1].
You can download the raw output, plot scripts and other material we used for our SIGMOD’18 paper here. Inside this zip file is another “data.zip” file, which contains the raw output. Our evaluation script “run.sh” does all the hard work of sifting through that data and extracting the information we needed for the AHEAD paper. Feel free to adapt that script to your needs. Keep in mind that the script automatically redirects the output to files and does log rotation for these output files (lower numbers are older!).
Please feel free to fork on github, send pull requests, and file issues in the issue tracker.
The Star Schema Benchmark and Augmented Fact Table Indexing
Nambiar R., Poess M. (eds) Performance Evaluation and Benchmarking. TPCTC 2009. Lecture Notes in Computer Science, vol 5895. Springer, Berlin, Heidelberg.
The SVG image for the link to the GitHub page is taken from GitHub.