Abstract
We describe a new deep learning approach to cardinality estimation. MSCN is a multi-set convolutional network, tailored to representing relational query plans, that employs set semantics to capture query features and true cardinalities. MSCN builds on sampling-based estimation, addressing its weaknesses when no sampled tuples qualify a predicate, and in capturing join-crossing correlations. Our evaluation of MSCN using a real-world dataset shows that deep learning signiicantly enhances the quality of cardinality estimation, which is the core problem in query optimization.
Original language | English |
---|---|
State | Published - 2019 |
Event | 9th Biennial Conference on Innovative Data Systems Research, CIDR 2019 - Pacific Grove, United States Duration: 13 Jan 2019 → 16 Jan 2019 |
Conference
Conference | 9th Biennial Conference on Innovative Data Systems Research, CIDR 2019 |
---|---|
Country/Territory | United States |
City | Pacific Grove |
Period | 13/01/19 → 16/01/19 |