Reproduce “RCRank: Multimodal Ranking of Root Causes of Slow Queries in Cloud Database Systems” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Biao Ouyang, Yingying Zhang, Hanyin Cheng, Yang Shu, Chenjuan Guo, Bin Yang, Qingsong Wen, Lunting Fan and 1 more

RCRank introduces a multimodal machine learning framework for diagnosing slow queries in cloud database systems, which identifies and ranks internal root causes based on their potential to accelerate query execution. The approach leverages SQL statements, execution plans, execution logs, and Key Performance Indicators, achieving superior performance over existing methods in both root cause identification and impact ranking, leading to an average runtime improvement of 24% to 30% for revised queries.

Key insights

  • Comprehensive slow query diagnosis requires "full observability" by integrating diverse multimodal data sources (SQL, plans, logs, KPIs) rather than relying on single-modal approaches.
  • Quantifying and ranking the potential impact of root causes on query acceleration is crucial for providing actionable insights, enabling users to prioritize efforts for maximum performance gain.
  • Sophisticated multimodal learning techniques, including self-supervised pre-training for cross-modal alignment and adaptive fusion mechanisms, are essential for effectively combining heterogeneous database-related data.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.