AutoSLO: Practical Latency SLOs on Cloud Data Warehouses
We present AutoSLO, a latency-SLO-aware workload management framework for multi-cluster cloud data warehouses. AutoSLO operates across three timescales through three key components.
The Policy Tuner can plan using the historical workload. It generates and simulates forecasted workloads to determine proactive cluster-management actions and configure parameters. By ensuring that the right resources are available at the right time (shortly before demand materializes), it reduces dependence on reactive scaling and mitigates cluster spinup delays.
The Autoscaler can adjust the active cluster set when observed workload behavior deviates from the forecast. Its important contributions include cluster spinup/teardown triggers, which identify opportune moments for resource adjustments, and a cluster spinup size selector, which uses short-term simulations to estimate the SLO and cost implications of different scaling actions.
Finally, the Query Router can react to live load. When a query q arrives, it selects on which active cluster it should be executed by managing a multi-way tradeoff among (1) the risk that q will violate its SLO; (2) the risk q poses to the SLO adherence of already-running queries; and (3) the resource cost implied by each routing decision.
-
Full API reference and documentation: mmarkakis.github.io/autoslo
-
Detailed map of paper concepts to repository files: Paper-to-code-mapping
-
Hands-on instructions on using our top-level scripts: Entry point reference.
Installation
To build the documentation site locally:
Cite This Work
@misc{markakis2026autoslo,
title={AutoSLO: Practical Latency SLOs on Cloud Data Warehouses -- Extended Version},
author={Markos Markakis and Tim Kraska},
year={2026},
eprint={2607.11770},
archivePrefix={arXiv},
primaryClass={cs.DB},
url={https://arxiv.org/abs/2607.11770},
doi={10.48550/arXiv.2607.11770}
}
A machine-readable citation is also available in CITATION.cff.