AXE: A Task Decomposition Approach to Learned LSM Tuning

Summary: AXE decomposes LSM tuning into learned cost modeling plus a tuner trained on unlimited synthetic samples, avoiding deployment-time retraining and handling categorical knobs. It scales across instances and environments, achieving higher performance than Bayesian Optimization 71% of the time with 100× lower overhead. (summarized by gpt-5.6-luna on Jul 24 2026)

Paper ID: 14393
Venue: VLDB
Year: 2025
Pagerank: 5.093636e-05
Overall Rank: 11,073 | 24.03%
DOI: 10.14778/3773731.3773735

Incoming Non-self Citations Over Time

No non-self incoming citations found for this paper in this database.

Authors

1. Andy Huynh (Boston University)
2. Anwesha Saha (Boston University)
3. Harshal A. Chaudhari (Boston University)
4. Manos Athanassoulis (Boston University)

BibTeX Citation

@article{huynh_vldb25,
        title = {{AXE: A Task Decomposition Approach to Learned LSM Tuning}},
        author = {Huynh, Andy and Saha, Anwesha and Chaudhari, Harshal A. and Athanassoulis, Manos},
        journal = {PVLDB},
        series = {{VLDB} '25},
        volume = {18},
        number = {13},
        pages = {5582--5595},
        doi = {10.14778/3773731.3773735},
        url = {https://doi.org/10.14778/3773731.3773735},
        year = {2025}
}

Incoming Citations (Sorted by Pagerank)

Showing 0 of 0 citing papers.

Rank	Citing Paper	Year	Venue	Pagerank

Outgoing Citations (Sorted by Pagerank)

Showing 30 of 30 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank	Cited Paper	Year	Venue	Pagerank
86	Automatic Database Management System Tuning Through Large-scale Machine Learning	2017	SIGMOD	0.00035316107
87	Automated Selection of Materialized Views and Indexes for SQL Databases	2000	VLDB	0.00035281619
156	An Efficient, Cost-Driven Index Selection Tool for Microsoft SQL Server	1997	VLDB	0.00028636811
259	Database Cracking	2007	CIDR	0.00023119313
334	An End-to-End Automatic Cloud Database Tuning System Using Deep Reinforcement Learning	2019	SIGMOD	0.00020875082
446	Monkey: Optimal Navigable Key-Value Store	2017	SIGMOD	0.00018332392
465	An End-to-End Learning-based Cost Estimator	2020	VLDB	0.0001803934
498	QTune: A Query-Aware Database Tuning System with Deep Reinforcement Learning	2019	VLDB	0.00017440583
831	Dostoevsky: Better Space-Time Trade-Offs for LSM-Tree Based Key-Value Stores via Adaptive Removal of Superfluous Merging	2018	SIGMOD	0.00013748512
1,015	AsterixDB: A Scalable, Open Source BDMS	2014	VLDB	0.00012647763
1,213	MyRocks: LSM-Tree Database Storage Engine Serving Facebook's Social Graph	2020	VLDB	0.00011646797
1,616	The Data Calculator*: Data Structure Design and Cost Synthesis from First Principles and Learned Cost Models	2018	SIGMOD	0.00010213691
1,942	Design Continuums and the Path Toward Self-Designing Key-Value Stores that Know and Learn	2019	CIDR	9.4451535e-05
2,740	Facilitating Database Tuning with Hyper-Parameter Optimization: A Comprehensive Experimental Evaluation	2022	VLDB	8.1855759e-05
2,822	Cost Models for Big Data Query Processing: Learning, Retrofitting, and Our Findings	2020	SIGMOD	8.0898536e-05
3,116	ResTune: Resource Oriented Tuning Boosted by Meta-Learning for Cloud Databases	2021	SIGMOD	7.7390737e-05
3,343	LlamaTune: Sample-Efficient DBMS Configuration Tuning	2022	VLDB	7.4983591e-05
3,577	Cosine: A Cloud-Cost Optimized Self-Designing Key-Value Storage Engine	2022	VLDB	7.2930211e-05
3,953	Efficient Deep Learning Pipelines for Accurate Cost Estimations Over Large Scale Query Workload	2021	SIGMOD	6.996368e-05
4,468	One Model to Rule them All: Towards Zero-Shot Learning for Databases	2022	CIDR	6.6819041e-05
4,489	COLT: Continuous On-Line Database Tuning	2006	SIGMOD	6.6675192e-05
5,766	Endure: A Robust Tuning Paradigm for LSM Trees Under Workload Uncertainty	2022	VLDB	6.094771e-05
5,780	Compactionary: A Dictionary for LSM Compactions	2022	SIGMOD	6.091058e-05
5,978	From Auto-tuning One Size Fits All to Self-designed and Learned Data-intensive Systems	2019	SIGMOD	6.0212877e-05
6,131	Foundations of Automated Database Tuning	2006	VLDB	5.966557e-05
6,843	Learning to Optimize LSM-trees: Towards A Reinforcement Learning based Key-Value Store for Dynamic Workloads	2023	SIGMOD	5.7573899e-05
6,921	Rethinking Learned Cost Models: Why Start from Scratch?	2023	SIGMOD	5.7388557e-05
7,757	CAMAL: Optimizing LSM-trees via Active Learning	2024	SIGMOD	5.5508469e-05
9,021	Structural Designs Meet Optimality: Exploring Optimized LSM-tree Structures in A Colossal Configuration Space	2024	SIGMOD	5.3305499e-05
9,257	MLOS in Action: Bridging the Gap Between Experimentation and Auto-Tuning in the Cloud	2024	VLDB	5.2972217e-05

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	8,360	SA-LSM: Optimize Data Layout for LSM-tree Based Storage using Survival Analysis	2022	VLDB
2	9,457	Rethinking The Compaction Policies in LSM-trees	2025	SIGMOD
3	3,343	LlamaTune: Sample-Efficient DBMS Configuration Tuning	2022	VLDB
4	2,970	Accordion: Better Memory Organization for LSM Key-Value Stores	2018	VLDB
5	5,229	Dissecting, Designing, and Optimizing LSM-based Data Stores	2022	SIGMOD
6	857	SlimDB: A Space-Efficient Key-Value Storage Engine For Semi-Sorted Data	2017	VLDB
7	7,757	CAMAL: Optimizing LSM-trees via Active Learning	2024	SIGMOD
8	5,766	Endure: A Robust Tuning Paradigm for LSM Trees Under Workload Uncertainty	2022	VLDB
9	5,226	Breaking Down Memory Walls: Adaptive Memory Management in LSM-based Storage Systems	2021	VLDB
10	7,459	Efficient Data Ingestion and Query Processing for LSM-Based Storage Systems	2019	VLDB