Skip to main content

Facebook's Presto Evolution into Trino and Starburst

Struggling with delivery, architectural drift, or platform instability?

I help teams fix systemic engineering issues across distributed systems, data platforms, and streaming pipelines.

→ See how I work with teams.


This retrospective revisits the early days of Facebook’s Presto engine in 2014, including installation quirks, missing security features, and benchmark comparisons with Hive and Tez. It explains how Presto’s performance and connector architecture reshaped SQL-on-Hadoop and ultimately led to the creation of Trino, the modern distributed SQL engine used today across large-scale data platforms.

From Presto to Trino: A Look Back at the Early Days of Distributed SQL

In late 2013, Facebook released Presto as an open-source distributed SQL engine. At the time, Hadoop’s dominant SQL engines—Hive (MapReduce), Hive+Tez, and early Impala—were still bound to batch-oriented execution models. Presto introduced something radically different: a low-latency, MPP-style SQL engine designed for interactive analytics at petabyte scale.

This article is a 2025 retrospective based on a hands-on write-up from 2014, preserving early installation notes and benchmark results while reflecting on how the technology evolved into Trino, one of the most important components in modern data platforms.

For additional background, see my related articles:

Presto in 2014: A New Approach to Interactive SQL

When Facebook open-sourced Presto, it immediately stood out:

  • Interactive SQL over Hive, HDFS, and Cassandra
  • An extensible connector architecture
  • A fully in-memory execution model (no MapReduce)
  • Petabyte-scale internal deployments at Facebook

Back then, Presto shipped with a minimal JDBC driver, early Hive connector support, and limited documentation—but it was already game-changing.

Installation in 2014: Simple, but with Caveats

Setting up Presto 0.68 required attention to a few sharp edges:

  • Java 7 only
  • No Kerberos support
  • No built-in authentication or authorization
  • Manual configuration of node.properties and config.properties

A typical setup looked like this:

uuidgen > node.id
datasources=jmx,hive
export PATH=/usr/jdk64/jdk1.7.0_45/bin:$PATH && presto-server-0.68/bin/launcher start

The Web UI showed splits, stages, and query execution paths—primitive compared to modern tooling, but revolutionary at the time.

The CLI was a standalone self-executing JAR:

export PATH=/usr/jdk64/jdk1.7.0_45/bin:$PATH && \
/software/presto --server host:port --catalog hive --schema default

Benchmarking Presto Against Hive and Tez (2014)

To evaluate performance, I generated sample datasets and executed a set of analytical queries. Here are two representative examples:

1. Finding Highest Gainers


select id, sum(amount) as amount 
from (
  select sender as id, amount * -1 as amount from transactions
  union all
  select recipient as id, amount from transactions
) t
group by id
order by amount desc
limit 10;

Results

Hive: 39.078 seconds
Tez:  18.227 seconds
Presto: 2 seconds

2. Finding Fraudsters (self-join with time window)


select count(*) 
from (
  select a.sender, a.recipient, b.recipient as c
  from transactions a 
  join transactions b on a.recipient = b.sender
  where a.time < b.time and b.time - a.time < 5
) t;

Results

Hive: 208 seconds
Tez:  101 seconds
Presto: 62 seconds

These benchmarks showed why Presto gained instant traction in large organizations: its interactive performance was often two orders of magnitude faster than Hive.

How Presto Evolved Into Trino

Over the next years, the Presto project diverged:

  • PrestoDB → the original Facebook-maintained version
  • PrestoSQL → spun out by the original creators

In 2020, PrestoSQL became Trino, now one of the most important engines in modern data platforms.

What Trino added:

  • Kerberos, LDAP, OAuth2 authentication
  • Fine-grained authorization (SQL grants, Ranger integration)
  • Iceberg, Delta Lake, Hudi connectors
  • Object storage support (S3, ADLS, GCS)
  • Fault-tolerant execution
  • Kubernetes-native deployments
  • Massive connector ecosystem

What started as a fast SQL engine for Hive tables became a universal analytics engine capable of querying dozens of data sources with ANSI SQL.

Modern Perspective (2025)

Looking back, Presto’s early design decisions reshaped an entire generation of data systems:

  • Interactive SQL became a baseline expectation
  • Federated analytics replaced ETL-heavy workflows
  • MPP execution became standard for analytical engines
  • The connector abstraction influenced modern lakehouse engines

The SQL-on-Hadoop ecosystem around Presto, Impala, Hive+Tez, Drill, and Tajo eventually gave way to the lakehouse ecosystem—Spark SQL, Trino, Flink SQL, Iceberg, Delta Lake, and cloud-native engines like BigQuery and Athena.

Presto may have started as a simple experimental engine running on Java 7, but it paved the way for the interactive analytical engines we rely on today.

Conclusion

Presto’s impact on the evolution of distributed SQL cannot be overstated. The 2014 benchmarks already showed the future: low-latency, MPP-style execution on open data formats. A decade later, this architecture has become the foundation of modern data platforms, with Trino standing as the direct descendant of the ideas introduced by early Presto.

Understanding this evolution helps modern data architects design systems that balance performance, openness, and ecosystem integration—principles that remain just as relevant today.


If platform instability, unclear ownership, or architecture drift are slowing your teams down, review my Services or book a 30-minute call.

Most read articles

Why Is Customer Obsession Disappearing?

Many companies trade real customer-obsession for automated, low-empathy support. Through examples from Coinbase, PayPal, GO Telecommunications and AT&T, this article shows how reliance on AI chatbots, outsourced call centers, and KPI-driven workflows erodes trust, NPS and customer retention. It argues that human-centric support—treating support as strategic investment instead of cost—is still a core growth engine in competitive markets. It's wild that even with all the cool tech we've got these days, like AI solving complex equations and doing business across time zones in a flash, so many companies are still struggling with the basics: taking care of their customers. The drama around Coinbase's customer support is a prime example of even tech giants messing up. And it's not just Coinbase — it's a big-picture issue for the whole industry. At some point, the idea of "customer obsession" got replaced with "customer automation," and no...

How to scale MySQL perfectly

When MySQL reaches its limits, scaling cannot rely on hardware alone. This article explains how strategic techniques such as caching, sharding and operational optimisation can drastically reduce load and improve application responsiveness. It outlines how in-memory systems like Redis or Memcached offload repeated reads, how horizontal sharding mechanisms distribute data for massive scale, and how tools such as Vitess, ProxySQL and HAProxy support routing, failover and cluster management. The summary also highlights essential practices including query tuning, indexing, replication and connection management. Together these approaches form a modern DevOps strategy that transforms MySQL from a single bottleneck into a resilient, scalable data layer able to grow with your application. When your MySQL database reaches its performance limits, vertical scaling through hardware upgrades provides a temporary solution. Long-term growth, though, requires a more comprehensive approach. This invo...

What the Heck is Superposition and Entanglement?

This post is about superposition and interference in simple, intuitive terms. It describes how quantum states combine, how probability amplitudes add, and why interference patterns appear in systems such as electrons, photons and waves. The goal is to give a clear, non mathematical understanding of how quantum behavior emerges from the rules of wave functions and measurement. If you’ve ever heard the words superposition or entanglement thrown around in conversations about quantum physics, you may have nodded politely while your brain quietly filed them away in the "too confusing to deal with" folder.  These aren't just theoretical quirks; they're the foundation of mind-bending tech like Google's latest quantum chip, the Willow with its 105 qubits. Superposition challenges our understanding of reality, suggesting that particles don't have definite states until observed. This principle is crucial in quantum technologies, enabling phenomena like quantum comp...