Datenschutzerklärung|Data Privacy

Martin Pagel

Mo 08.06.2020, 16:00 - 17:00 Uhr, Online: "Cardinality estimation in modern DBMS and through sampling-based approaches" (Johannes Maeß, DHBW Stuttgart)

Cardinality estimation has been the subject of publications for decades and is still seen as a major unsolved problem in query optimization research.
Modern DBMS estimate cardinalities based on incorrect assumptions of uniformity and independence of the underlying data. Such heuristics commonly yield significant errors in cardinality estimates and hence negatively impact query performance.

This talk analyses the cardinality estimator of a commercial DBMS and discusses how these assumptions fail in complex database schemas and queries:
How accurate are the estimates across different types of table relationships? How degenerating are uneven distributions of join predicates? Where do errors get introduced, and how do they propagate through complex query paths?

This talk goes into detail about the limits of database-internal statistics for cardinality estimation and how more advanced techniques like sampling can improve intermediate and final cardinality estimates.

Johannes Maeß is a third-year Bachelors's student at DHBW Stuttgart studying Business Information Systems. He obtained research experience in the field of databases and, in particular, query optimization as an intern at the IBM Almaden Research Center, San Jose.

Additional experience includes internships in different technical roles in the IBM corporate studies program; his further interests lie in Machine Learning and its applications in Computer Vision and NLP.

Log-in information:
If you are interested in attending the online presentation, please contact!