Performance
Take performance to the next level with Starburst.
Trino was built from the ground up to run SQL queries fast and at scale. It was built to solve the low performance, and lack of connectivity associated with using Hive. The creators of the Trino (formerly known as Presto SQL) project, needed to handle hundreds of users issuing tens of thousands of queries each day against petabytes of data. Trino was built for performance, on any data source, anywhere.
Starburst has made a series of enhancement to open source Trino to greatly improve query performance. First, Starburst provides improved and exclusive connectors which expose statistics allowing the query engine to generate faster and more optimal query plans. Starburst connectors also can pushdown the processing of queries, or parts of queries, into the connected data source to improve query performance. Dynamic filtering optimizations significantly reduces network traffic and load on data sources. And Starburst Cached Views include a suite of performance upgrades, with different options for accessing frequently used data.
Make better decisions faster than ever
Parquet reader
Trino has long served organizations with the ability to read Parquet files. And while already fast, Starburst delivers additional value with the Starburst Parquet reader which improves read performance on Parquet by an average of 20% over Trino.
Cost-based optimizer
Depending on the complexity of your SQL query there are many, often exponential, query plans that return the same result. However, the performance of each plan can vary drastically; taking only seconds to finish or days given the chosen plan. The cost-based optimizer creates the optimal path for each possible query plan. Starburst enables the cost-based optimizer to be leveraged across the majority of connectors.
View Documentation
Starburst Cached Views
Starburst Cached Views is a collection of performance features that provides customers with options on accessing frequently used data. Starburst customers benefit from materialized views, table scan redirections, and cache service/CLI for easy management, configuration, and most importantly, speed for accessing the same, or similar, data sets repeatedly. More efficiency, less work.
Pushdown
Sometimes it’s better to lean on other systems. Starburst can pushdown the processing of queries, or parts of queries, into the connected data sources resulting in better query performance, reduced network traffic and decreased load on the data source.
View Documentation
Dynamic filtering
Dynamic filtering optimizations significantly improves the performance of queries with selective joins by avoiding the read of data that would be filtered by join condition. It significantly boosts the performance of query federation.
View Documentation
