STARBURST ENTERPRISE PRESTO:

NATIVE DELTA LAKE READER

The secure, enterprise-grade distribution of the open

source Presto SQL query engine now includes a Native

Delta Lake Reader.

Open-sourced by Databricks in 2019, Delta Lake enables data modication

and optimizations in data lakes. The Native Delta Lake Reader helps Databricks

customers take advantage of Presto’s speed, concurrency, and scalability to query

their Delta Lake. Starburst and Databricks share many of the same enterprise

forced to modify, join, and overwrite tables. Databricks changed this with its Delta

Lake, a storage platform that lets users easily update and modify data stored

in a cloud data lake. Additionally, Delta Lake provides performance and le

management optimizations which didn’t exist in cloud data lakes.

Starburst Native Delta Lake Reader

On a mission to power analytics anywhere, Starburst recognized the need to

support the leading technology that enables ACID transactions and performance

optimizations on top of object storage. Starburst’s Native Delta Lake Reader was

written from scratch, specically for Delta Lake, to make it as efcient as possible.

Features include:

Starburst always supports the best platforms rst — and Delta Lake is the industry

leader and a favorite tool of some of our top enterprise customers.

DATASHEET

Starburst Enterprise Presto: Native Delta Lake Reader

Starburst & Databricks: Complementary Platforms

Databricks and Starburst share many of the same customers, and these large global enterprises have been asking for

a Native Delta Lake Reader. Our shared customers overturn a common misconception in the industry — the idea that

Databricks and Starburst are competitors. Instead, the two are complementary. Enterprises use Spark and Databricks

for Machine Learning, AI, ETL, and streaming ingestion, while Starburst Enterprise Presto is their high-concurrency SQL

Delta Lake, these les are stored in an open-

source format called Parquet. In Delta Lake, each

time there is a change to a le — when customer

information is modied, for example — this is added

to the transaction log, along with an associated

timestamp. Starburst’s new tool reads the Delta

Log and les (when needed) and extends multiple

benets to end users.

Data Sources

Other Data Sources

Data Lake Machine Learning / AI

ON-PREMISES

Streaming

Batch

Previously you’d have to rewrite an entire table to add a

new customer address or some other additional piece

simple as a delete or update statement. As part of our

commitment to ne-grained global security, Starburst’s

Native Delta Lake Reader supports this functionality.

Min, max, nulls, counts and other high-level statistics

are stored on each le. Starburst took advantage of this

when designing the Native Delta Lake Reader. Users can

quickly narrow which les they actually need to query.

The Starburst connector feeds this information into the

Presto Cost Based Optimizer, which greatly improves

The process of updating data does leave you with many

small les, and although object storage can be pretty

command that combines small les into larger ones,

which greatly increases performance. Starburst’s Native

Delta Lake Reader allows users to take advantage of

this as well.

After you optimize les, and are left with, say, 10,000

les instead of 100,000, you can order them by the

columns of your choosing. Included in the optimize

command, you can optionally choose to z-order

these les by one or more columns. This increases

The Reader allows you to:

•

Create tables in Starburst Presto that point to

Delta Lake tables

•

Execute queries on Delta Lake data as you would with

normal tables

•

Optimize and z-order your tables to improve performance

All of this is seamless to the end user. Overall, the Native