Learned Indexes for a Google-scale Disk-based Database
arXiv (Cornell University)2020
Citations Over Time
Hussam Abu-Libdeh, Deniz Altınbüken, Alex Beutel, Ed H., Lyric Doshi, Tim Kraska, Xiaozhou Li, Li, Andy Ly, Olston, Christopher
Abstract
There is great excitement about learned index structures, but understandable skepticism about the practicality of a new method uprooting decades of research on B-Trees. In this paper, we work to remove some of that uncertainty by demonstrating how a learned index can be integrated in a distributed, disk-based database system: Google's Bigtable. We detail several design decisions we made to integrate learned indexes in Bigtable. Our results show that integrating learned index significantly improves the end-to-end read latency and throughput for Bigtable.
Related Papers
- → Expert Oracle Indexing and Access Paths(2016)18 cited
- → Parallel multi-tree indexing for evaluating large descriptor sets(2013)5 cited
- Indexing the past, present and future positions of moving objects using PPFI*(2012)
- → Multiple Entry Indexing and Double Indexing(2007)1 cited
- Research on Time Sensitive Information Retrieval Based on Real Time Search Engines Model(2014)