Excited to share that we got three papers recently accepted at VLDB and EDBT that focus on delivering better indexing! Pre-prints are available below:
- Our journey to developing sortedness-aware indexing continues with a simple yet very powerful design called Quick Insertion Tree (QuIT), which allows near-bulk-load index ingestion when data comes near-sorted without any performance hit on read queries! "QuIT your B+-tree for the Quick Insertion Tree" will appear in EDBT 2025, Barcelona.
- We also have exciting results to share about our goal to transform bitmap indexes to general-purpose indexes capable of handling (almost) any workload (including update-heavy with high concurrency). We developed an approach for Concurrent Updatable Bitmap Indexing (or CUBIT) that uses lock-free fine-granular out-of-place updates to offer superb performance and scalability. "CUBIT: Concurrent Updatable Bitmap Indexing" will appear at VLDB 2025, London.
- Finally, our experimental analysis on Write Amplification in RocksDB helped us uncover a variation of the round-robin file-picking policy for compaction that is very close to the best possible design, both in terms of performance stability and write amplification. "Benchmarking, Analyzing, and Optimizing Write Amplification of Partial Compaction in RocksDB" will appear at EDBT 2025, Barcelona.
