Andy Pavlo
Professor of Computer Science · Carnegie Mellon University · Distinguished
Andy Pavlo is a Professor of Computer Science at Carnegie Mellon University, known for the self-driving database systems Peloton and NoisePage.
- lectures
- 47
- series
- 3
- hours
- 60
Start here
Series and courses
CMU 15-799: Query Optimization in Modern Database Systems
Open the course
- 1SQL Optimization Background
- 2History of Query Optimizers ft. IBM System R
- 3IBM Starburst Query Rewriter + Optimizer
- 4EXODUS + Volcano Query Optimizer Generators
Show 10 moreShow fewer
- 5Cascades Query Optimizer
- 6Query Plan Transformations
- 7Join Ordering: Bottom-up
- 8Join Ordering: Top-Down / Transformational
- 9Search Parallelization: Bottom-up
- 10Search Parallelization: Top-Down
- 11Unnesting Queries
- 12Query Cost Models: Statistics
- 13Query Cost Models: Cardinality Estimation
- 14Query Cost Models: More Cardinality Estimation
CMU 15-721: Advanced Database Systems
Open the course
- 1Modern OLAP Database Systems
- 2Data Formats & Encoding Part 1
- 3Data Formats & Encoding Part 2
- 4Query Execution & Processing Part 1
Show 17 moreShow fewer
- 5Query Execution & Processing Part 2
- 6Vectorized Query Execution Using SIMD
- 7JIT Query Compilation & Code Generation
- 8Query Scheduling & Coordination
- 9Parallel Hash Join Algorithms
- 10Multi-Way Join Algorithms / Worst-Case Optimal Joins
- 11User-Defined Function Optimizations
- 12Database Networking Protocols
- 13Query Optimizer Implementation 1
- 14Query Optimizer Implementation 2
- 15Query Optimizer Implementation 3
- 16Google BigQuery / Dremel
- 17Databricks Photon / Spark SQL
- 18Snowflake Data Warehouse Internals
- 19DuckDB Embedded Database System
- 20Yellowbrick Data Warehouse System
- 21Amazon Redshift Data Warehouse System
CMU 15-445/645: Intro to Database Systems
Open the course
- 1The Relational Model & Relational Algebra
- 2Modern SQL — Window Functions, CTEs & Bag Semantics
- 3Database Storage I — Volatile vs. Non-Volatile Hierarchy & Slotted Pages
- 4Database Storage II — Log-Structured Merge Trees (LSM) & Tuple Layout
Show 8 moreShow fewer
- 5OLTP vs. OLAP Storage Models & Database Compression
- 6Buffer Pool Managers, LRU-K Eviction & Direct I/O
- 7Static & Dynamic Hash Tables (Linear Probe, Robin Hood, Cuckoo, Extendible)
- 8B+Tree Indexes
- 9Index Concurrency Control
- 21Intro to Distributed Databases
- 22Distributed Transaction Processing Databases
- 23Distributed Data Warehouse OLAP Databases
In these lectures
- Declarative SQL separates the logical intent of a query from its physical execution, allowing database systems to adapt to changing data.History of Query Optimizers ft. IBM System R
- Traditional optimizers generate a fixed plan before execution based on static assumptions, which often fail due to inaccurate cardinality estimations or changes in database state.Query Optimizer Implementation 3
- Cost models rely on separating physical hardware costs from logical intermediate result sizes.Query Cost Models: Statistics
Draws on
People cited in the lectures, at the moment of the citation.
- Thomas NeumannResearcher cited on the logical query optimization slide footer
- Michael StonebrakerDatabase researcher cited for commentary on Oracle's early optimizer.
- Jim GrayTuring Award winner for database work in the 1980s.
- Goetz GraefeInventor of Volcano and Cascades who won the ACM SIGMOD Test of Time Award.
- Peter BonczResearcher associated with CWI and database systems.
- Guy LohmanResearcher and author of blog articles on DB2 cost models.
Playlists
1About
Andy Pavlo is a Professor of Computer Science at Carnegie Mellon University, known for the self-driving database systems Peloton and NoisePage. Graduate lectures from CMU 15-721 and 15-799 cover System R cardinality estimation, subquery unnesting, and the internals of Snowflake and DuckDB.
Taught with
Same subject, Computer Systems & Software
Connections
55Their places, courses, books and lectures.
Courses3
Concept1
Lectures47
- History of Query Optimizers ft. IBM System R
- Query Optimizer Implementation 3
- Query Cost Models: Statistics
- User-Defined Function Optimizations
- Modern OLAP Database Systems
- SQL Optimization Background
Show 41 moreShow fewer
- Database Networking Protocols
- Snowflake Data Warehouse Internals
- Query Optimizer Implementation 2
- IBM Starburst Query Rewriter + Optimizer
- DuckDB Embedded Database System
- Query Cost Models: Cardinality Estimation
- Data Formats & Encoding Part 1
- Join Ordering: Bottom-up
- Query Plan Transformations
- EXODUS + Volcano Query Optimizer Generators
- Google BigQuery / Dremel
- Databricks Photon / Spark SQL
- Unnesting Queries
- B+Tree Indexes
- Cascades Query Optimizer
- Index Concurrency Control
- Query Scheduling & Coordination
- Yellowbrick Data Warehouse System
- Multi-Way Join Algorithms / Worst-Case Optimal Joins
- Query Cost Models: More Cardinality Estimation
- Query Optimizer Implementation 1
- Query Execution & Processing Part 2
- Query Execution & Processing Part 1
- Parallel Hash Join Algorithms
- JIT Query Compilation & Code Generation
- Join Ordering: Top-Down / Transformational
- Search Parallelization: Top-Down
- Vectorized Query Execution Using SIMD
- Amazon Redshift Data Warehouse System
- Static & Dynamic Hash Tables (Linear Probe, Robin Hood, Cuckoo, Extendible)
- Database Storage II — Log-Structured Merge Trees (LSM) & Tuple Layout
- Intro to Distributed Databases
- Data Formats & Encoding Part 2
- Search Parallelization: Bottom-up
- Distributed Data Warehouse OLAP Databases
- Distributed Transaction Processing Databases
- Database Storage I — Volatile vs. Non-Volatile Hierarchy & Slotted Pages
- The Relational Model & Relational Algebra
- Buffer Pool Managers, LRU-K Eviction & Direct I/O
- OLTP vs. OLAP Storage Models & Database Compression
- Modern SQL — Window Functions, CTEs & Bag Semantics






