Advances in Query Optimization and Data Analytics

The field of database management and data analytics is moving towards more efficient and adaptive query optimization techniques. Recent developments focus on leveraging machine learning and data-agnostic approaches to improve query performance and reduce training costs. Noteworthy papers in this area include Delta, a mixed cost-based query optimization framework that achieves an average 2.34x speedup over PostgreSQL, and GRASP, a data-agnostic cardinality learning system that operates without data access and uses only 10% of all possible join templates. Additionally, researchers are exploring novel methods for workload synthesis, such as PBench, which reduces approximation error by up to 6x compared to state-of-the-art methods. Other notable advancements include the development of high-performance GPU implementations for constructing H2 matrices and the creation of adaptive sketching algorithms for online analytics.

Sources

Delta: A Learned Mixed Cost-based Query Optimization Framework

Data-Agnostic Cardinality Learning from Imperfect Workloads

PBench: Workload Synthesizer with Real Statistics for Cloud Analytics Benchmarking

Adaptive Sketching Based Construction of H2 Matrices on GPUs

LMQ-Sketch: Lagom Multi-Query Sketch for High-Rate Online Analytics

Towards an Introspective Dynamic Model of Globally Distributed Computing Infrastructures

ClusterRCA: Network Failure Diagnosis in HPC Systems Using Multimodal Data

Scalable GPU Performance Variability Analysis framework

Built with on top of