The document provides an overview of Trino, a fast distributed SQL query engine designed for large datasets across various sources, highlighting its evolution from Presto and integration with Apache Pulsar. It discusses the advantages of Trino, including its query federation capabilities, and the development history along with community engagement. The document also includes links to community resources and contributions related to Trino.

1. Trino: A ludicrously fast query engine
Brian Olsen
Developer Advocate @
Trino Contributor
@bitsondatadev

2. Overview
● Pulsar SQL
● How Pulsar uses Presto
● Trino Overview
○ Hive: The perfect SQL HiveQL solution
○ Presto! Your queries are fast!
○ Trino (formerly known as PrestoSQL)
● Query Federation with Pulsar
● Demo
● Looking Forward

3. Pulsar SQL
● Pulsar is message/streaming or
event store?
● Segment centric storage in Apache
Bookkeeper
● Historical data (old segments) that
can exist in bookies or in S3
● Jerry Peng added PIP-19 Aug 2018
● Adds Pulsar SQL
● https://github.com/apache/pulsar/pull/2265

4. How Pulsar uses Presto
● Pulsar stores schema in registry
stored in Apache bookkeeper
● Uses Avro definition for schemas
● Provides structured streams that
allow data to be represented in
tabular format (ANSI compliant)

5. How Pulsar uses Presto...or is it Trino?
● Uses embedded Presto and
its own scripts to manage the
cluster
● March 2020: Issue 6605
prestodb changes to prestosql
● July 2020: Yuya Ebhyr
updates to prestosql v332
● Dec 2020: Trino rebrand from
prestosql
● June 2021: Pulsar community
donates Trino-Pulsar
connector (Marvin Cai) to
Trino

6. Presto vs Trino
● Wait isn’t today’s talk about Trino?
● Why are we talking about Presto?

7. Trino is a fast distributed SQL query engine designed to query large data sets
distributed over one or more heterogeneous data sources.
Trino
Trino TL;DL
Trino Clients Data Sources

8. Hive: The perfect SQL HiveQL solution
Developers at Facebook created Hive, a SQL-on-Hadoop solution that takes a SQL
like syntax, HiveQL, and transforms it into MapReduce operations on data in
Hadoop.
● Simplified development process
● Queries still taking long
● Established a base for how to model
SQL tables of generic data stores

9. Presto! Your queries are fast!
Martin Traverso, Dain Sundstrom, and David Phillips created Presto in 2012. It
aimed to solve for the slow queries of Hive at Facebook and eventually many more.
Development Philosophies:
● Open source with neutral governance model
● It just works (Netezza was a commercial inspiration)
● Fast, interactive analytics
● Correct results
● Standards based (ANSI SQL, JDBC, etc..)
Facebook management unilaterally rewrites
rules around committership in late 2018

10. Trino (formerly known as PrestoSQL)
● If you know Presto or are using PrestoSQL…
○ same software
○ same people
○ same community
○ under a shiny new name
○ and a cute bunny
● Companies don’t run open source projects,
people do.
● More details:
https://trino.io/blog/2020/12/27/announcing-trino.html

11. Query Federation with Pulsar
Facebook specific queries Open source queries

12. Query Federation with Pulsar
SPI
Facebook specific connectors Open source connectors

13. Query Federation with Pulsar (Pulsar Centric)
SPI

14. Query Federation with Pulsar (Trino Centric)
SPI

15. Demo

16. Looking Forward

17. Community
● Trino Slack
○ https://trino.io/slack.html
● Trino Community Broadcast
○ https://trino.io/broadcast/
● Trino Virtual Meetups
○ https://www.meetup.com/trino-americas
○ https://www.meetup.com/trino-emea
○ https://www.meetup.com/trino-apac
● Contribute to the project
○ https://trino.io/development/
● Write blogs or docs
○ https://trino.io/blog/
○ https://github.com/trinodb/trino/tree/mas
ter/docs

18. Thank you
Please give us a ⭐ on
github.com/trinodb/trino
@bitsondatadev