SmokedDuck: Lineage via Selection Vector Capture


Video


Team Information

Team Members

  • Charlie Summers, DES Candidate in Computer Science, Columbia Engineering

  • Haneen Mohammed, PhD in Computer Science, Columbia Engineering

  • Faculty Advisor: Eugene Wu, Associate Professor of Computer Science, Columbia Engineering

Abstract

We introduce Selection Vector Capture - a new technique providing blazing fast lineage capture in vectorized databases by pinning existing data structures in memory. We showcase these techniques in SmokedDuck, a fork of the open-source database DuckDB.

Team Lead Contact

Charlie Summers: cgs2161@columbia.edu

Previous
Previous

Should Personalization Be Optional on Paid Streaming Platforms?: An Experiment on User Preferences for Personalization or Increased Data Privacy

Next
Next

Interactive Visualization Interface Generation for SQL Analysis in Notebook