[Experiments & Analysis] Evaluating Learned Indexes in LSM-tree Systems: Benchmarks, Insights and Design Choices DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation emai; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/18/06
Junfeng Liu, Jiarui Ye, Mengshi Chen, Meng Li, Siqiang Luo · 2025
Read [Experiments & Analysis] Evaluating Learned Indexes in LSM-tree Systems: Benchmarks, Insights and Design Choices DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation emai; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/18/06 by Junfeng Liu, Jiarui Ye, Mengshi Chen, Meng Li, Siqiang Luo (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-08671.txt.
Description
Junfeng Liu and four co-authors benchmark learned indexes, the machine-learned lookup structures now being grafted onto LSM-tree databases. Their 2025 paper notes that these stores power much of industry, but lookups strain as data grows. Past studies tried only a few learned index types, leaving their trade-offs murky.
The team map the workflow of eight learned indexes and cost them out in theory. They define a design space covering index type, boundary placement, and granularity, then build all variants on one platform for fair tests. Surprises emerged: pouring memory into learned indexes barely helped lookups, and retraining cost less than feared. The authors close with practical tuning guidance for developers choosing among the designs.
Catalog record
| CREATOR | Junfeng Liu, Jiarui Ye, Mengshi Chen, Meng Li, Siqiang Luo |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-08671