Analytics Anywhere

About Starburst Our Platform

600% Growth YoY Named Open Source ANSI SQL MPP On-Prem, High Massive
Startup to Watch Query Engine or Cloud Concurrency Scale
2020

82

100+ NPS Score Rapid Time to Low Cost of Enterprise 24x7 Expert
Enterprise Insights Ownership Grade Security Support
Customers

2
 Our Customers Our Value to Them

Reduced
Decision Risk

Increased Revenue
and Profit

Higher Customer
Retention

Accelerated Time
to Market

3
 Today’s data management approach delays analytics
Business has a question Data Engineering services the request Business gets an answer

ETL Data Data Lake

Warehouses

Database

Multiple Copies
Cloud Data Cloud Data
Warehouse Lake

Delays decision-making, increases data costs & complexity

4
Comcast Journey

18 Months 5 Weeks 0 Disruption 93% Faster

Existing Hadoop $200M in Migrating to S3 Adding Delta Lake
Data Lake too Revenue from Teradata and Kafka
TimelineSlow
provided by Timeline to execute the Accelerate cloud Begin ingesting new data
team to execute CMO same CMO cross sell migration without sources in real time for
cross sell campaign campaign business disruption predictive analytics

5
Connectivity: Creating a Portable Access Layer
Data Scientists Finance Marketers Data Analysts

Existing analytics tools

The Data Consumption Layer

Fine-grained access Column + Row-level

Data Masking Data Encryption Query Auditing Global Security control permissions

Data Lakes Relational Databases NoSQL Stores Publish/Subscribe

Azure Event Hub

6
 Starburst Trino: SQL Engine Architecture
Data: Storage
Trino Cluster: Compute
Report,
SQL Coordinator Parse
Node Metadata
API Glue/Hive
Optimize
Catalog
Results (CBO)
Data Location
BI Tool, SQL Client, API
CLI Schedule

Key Data
Worker
Coordinator Node GCS ADLS Blob Storage S3

Worker Worker
Node
Auto-scaling group
Data
Worker
Connectors
Node
ODBC/JDBC, CLI
Intra-Cluster
API Call
Separation of compute and storage
Deploy Starburst everywhere - On-premise or Cloud

Hive Metastore
Horizontal Pod
via Helm Charts Service Object Store
Autoscaler (HPA) Storage
Pod

Hadoop / Hive
/ Delta

Starburst Starburst Admin Starburst Worker

Service Presto Worker
Pod Presto Worker Any RDBMS
Pod
Pod Config Properties:
connection-url=<pgsql_url>
connection-password=<password
>
Starburst connection-user=<sb-user>
RDS=PostgreSQL
Coordinator
Pod
Hybrid Deployment - Standard Approach
Global Users
Local Users

● High Latency for large

datasets dependent on the
network connectivity

● High Costs generated by

data movement required for
each query
Data Read
● Metadata duplicated in
Data Read
different environments

● No Control around Data

Residency

Cloud Storage Local Storage

Hybrid Deployment - Starburst Approach
Global users Local Users

● Low Query Latency with reduced

network traffic

Starburst-Remote
● Fully scalable approach that
Connector allows connection between all
your environments

● Unified Metadata model with no

Data Read data duplication
Data Read

● Improved Security Model that

meets any data residency
requirements

Local Storage
Cloud Storage
A Federated Semantic Layer - powered by Trino ( formerly Presto SQL )

● Access data in real-time -

where it lies

● Connect the tool of your

choice

● Different clusters for

different functions
(chargeback)

● Build business views over a

variety of sources

● Additional access control

over all sources
 Don’t just take our word for it…
Faster analytics on data in your data lake Faster analytics on data anywhere + query federation

Data Engineers Data Engineers

• “This streamlined workflow helps our executives make

• 99.9% cluster uptime to drive higher CEX
the right decisions on time, and fosters innovation
• 3x performance improvement over OS Presto through machine learning.”

• 25% reduction in TCO • 50% savings on infrastructure compute costs to

expand use cases with same budget

Data Engineers

• Achieved GDPR compliance by leaving data where it lives • Eliminated ETL for joining Oracle, HDFS while
locally replacing Spark/Impala

• Reduced infrastructure usage by 30% • Reduce time to insight for critical risk models 96%

• Improved time to insights for engineers/analysts by 800% • De-risk business decisions in real time

12