1. Confidential – Oracle Internal/Restricted/Highly Restricted

2. Big Data Made Simple with Oracle Data Flow
Just Add Code
Big Data Made Simple with
Oracle Data Flow
Carter Shanklin Product Management Oracle
Confidential – Oracle Internal/Restricted/Highly Restricted

4. The Challenge of Big Data
Flexible frameworks for ETL, SQL, ML and more.
Process huge datasets.
Retain data longer than ever.
Extreme complexity.
Lack of skilled operations teams.
Difficult to align cost and value.
Over the past 5 or so years, Big Data has led to amazing advancements across all industries.
My personal favorite Big Data story is British Airway’s Know Me program, which by combining data across disparate inputs, is able to do things like predict when a customer is going to miss a flight and automatically offer to re-book them on another flight.
There are hundreds of stories of how Big Data has changed the business landscape by bringing in data from many fronts, combining them and analyzing them.
Everyone understand the potential of Big Data, but most people have not realized it. Why?
Confidential – Oracle Internal/Restricted/Highly Restricted

5. Big Data Is Just Too Complicated!
“70% of Hadoop deployments will fail to meet cost savings and revenue generation objectives due to skills and integration challenges.”
- Merv Adrian, VP Research, Gartner
“For three successive years the number of deployments for (Hadoop) projects for our clients has been about 15%. It's hardly moved.”
- Merv Adrian, VP Research, Gartner
Gartner has looked at this question in great detail, and reports that only about 15% of Big Data projects make it into production.
The reason is complexity. Big Data infrastructures require deep operational expertise that is hard to find, hard to hire and hard to retain, so as much as companies want to use Big Data, they don’t have the skills and ability to do it.
We need a better way to consume Big Data.
Source: get-slow-start-in-enterprise/d/d-id/ ?print=yes
March 20, 2018
“...technologies such as Hadoop and Spark are new, they are difficult, and they are complicated.”
- Merv Adrian, VP Research, Gartner
Confidential – Oracle Internal/Restricted/Highly Restricted

6. Oracle Data Flow Makes Big Data Truly Easy1
Fast: Launch Apache
Spark® Jobs in Just
Seconds. 2
Serverless: No
Infrastructure to
Deploy or Manage. 3
Complete: Process
data in the Cloud or
in your Datacenter.
Realizing that complexity is the main challenge, we have begun development on Oracle Data Flow, which lets you run Apache Spark® jobs in a true Serverless model
Spark as you may know has emerged as the leading Big Data processing framework, supporting SQL, ETL, Graph, ML and more, and is at the center of Modern Big Data.
Data Flow focuses squarely on solving the main complexity challenge we all face with Big Data. With no infrastructure to manage, no OS to patch or tune, nothing to upgrade, the most complex aspects of Big Data are handled for you transparently and automatically, allowing you to focus on the analytics, focus on those analytics that make it so your customers just can’t live without you.
On top of this, Data Flow launches almost instantly, which helps you stay agile whether you’re in test or dev
Data Flow Supports any type of Spark job including SQL, Python, Java, Scala and R.
And Data Flow can be Run via UI or API, for easy integration with your production pipelines or your favorite scheduler.
Oracle Data Flow
The easiest way to run Apache Spark® in the cloud.
Serverless, so you never worry about any infrastructure.
Any type of Spark job, no re-coding required.
Access data from any source, wherever it lives.
Availability targeted for 2019.
Oracle Confidential – Restricted

7. Oracle Data Flow: Fast Launch1
Fast: Launch Apache
Spark® Jobs in Just
Seconds. 2
Serverless: No
Infrastructure to
Deploy or Manage. 3
Complete: Process
data in the Cloud or
in your Datacenter.
Let’s drill down into some of these points before we jump into a demo
Oracle Data Flow
Launch in Seconds,
Simple UI and API
Hadoop on Prem:
Weeks to Months
Hadoop in Cloud:
20 Minutes or More
Oracle Confidential – Restricted

8. Oracle Data Flow: Serverless1
Fast: Launch Apache
Spark® Jobs in Just
Seconds. 2
Serverless: No
Infrastructure to
Deploy or Manage. 3
Complete: Process
data in the Cloud or
in your Datacenter.
Upgrading and re-certifying Hadoop can take weeks to months.
HDFS upgrades touch all data and can lose/corrupt data
Hadoop APIs are not backward compatible and frequently break apps
YARN and HDFS performance characteristics differ substantially between versions, affecting post- upgrade SLAs
Serverless lets you choose a compatible version for each job.
No data upgrade.
Roll development forward independently of production
Application performance is isolated
Net result is you can upgrade in hours/days rather than weeks/months
Hadoop: 30% or more of your time spent keeping the lights on.
Oracle Data Flow: No patches, easy to upgrade = spend 100% on innovation.
MAY s m t w f 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
MAY s m t w f 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
Dev: Spark 2.4
Prod: Spark 2.3
Downtime: Upgrade OS
Downtime: Upgrade Hadoop
Switch to Spark 2.4
Certify New Hadoop
Oracle Confidential – Restricted

9. Oracle Data Flow: Complete1
Fast: Launch Apache
Spark® Jobs in Just
Seconds. 2
Serverless: No
Infrastructure to
Deploy or Manage. 3
Complete: Process
data in the Cloud or
in your Datacenter.
Oracle Cloud Data Lake
Oracle Object Storage
SaaS
Oracle Autonomous DB
Co-Located
Access
Oracle Data Flow
Compute
Multi Cloud
Oracle Data Flow
S3 / WASB
NoSQL
RDBMS
Compute
Oracle Data Flow
Hybrid Cloud
Compute
Object Storage
EDW
Private
VPN
Oracle Catalog

10. How Does it Work? Run Any Spark Job in 3 Steps1 2 3
Sign Up
Upload your Spark App
Run!
.py
.jar
SQL
Data in Oracle Object Storage
cloud.oracle.com
Includes $300 of free
trial credits
Configure your app
Select number of VMs
Run!
Oracle Confidential – Restricted

11. Oracle Data Flow: Access Data Wherever it Lives1 2 3
In Oracle Cloud
From Other Clouds
From Your Datacenter
Secure via VPN
1: Includes all the connectors you need to talk to Database or MySQL
2. Connect to any cloud to offload analytics
3. Peer a virtual private network to access your on -premises data for data that doesn’t live in the cloud
Azure
AWS
Oracle
Oracle Data Flow
Oracle Autonomous Database
MySQL
Oracle Object Storage
Oracle Data Flow
Private VCN
On-Prem EDW
Oracle Confidential – Restricted

12. Oracle Data Flow: Secure Execution
Control
Your Spark jobs are isolated in private VMs.
Data is always encrypted, at- rest and in-motion.
Service Controller
dfcs.oraclecloud.com
Customer 1 Job
Customer 2 Job
Spark Driver
Spark Executor
Spark Executor
Spark Driver
Spark Executor
Spark Executor
Spark Executor
Spark Executor
Storage Container
Storage Container
Customer 2 Objects
Customer 2 Objects
Customer 2 Objects
Oracle Confidential – Restricted

13. Oracle Data Flow: No Complex Capacity Planning
Control
Traditional Model: Noisy neighbors mean you provision to maximum expected load.
Data Flow: Jobs run on VMs that come and go on-demand.
No noisy, no expensive max-load provisioning.
Hit SLAs while paying for only what you need.
Service Controller
dfcs.oraclecloud.com
Customer 1 Job
Customer 2 Job
One of the most complex problems is sizing shared clusters to ensure predictable, consistent SLAs, when workloads are bursty and when workloads come and go.
In traditional big data architectures you have to size your cluster to handle the busiest possible load, meaning you massively overspend on your big data architectures
Spark Driver
Spark Executor
Spark Executor
Spark Driver
Spark Executor
Spark Executor
Spark Executor
Spark Executor
Storage Container
Storage Container
Customer 2 Objects
Customer 2 Objects
Customer 2 Objects
Oracle Confidential – Restricted

14. Oracle Data Flow is API-Driven.
Service driven entirely by REST APIs.
Integrate with your applications or workflow engine of choice.
Oracle Object Storage
Oracle Data Flow
Spark Job
Spark Jobs
REST APIs
Secure
Read / Write
Oracle Confidential – Restricted

15. Big Data Made Really Simple
DEMO
Oracle Data Flow
Big Data Made Really Simple
Confidential – Oracle Internal/Restricted/Highly Restricted

16. Oracle Data Flow Demo: Convert XML Mess into a SQL Table for Reporting
Documents
Parquet Table
Oracle Data Flow
Apache Spark Processing
Fully Serverless
Oracle Object Storage
Oracle Compute
Oracle Object Storage
Oracle Confidential – Restricted

17. Summary Oracle Data Flow Simplest Way to run Spark in the Cloud.
Any type of Spark job, any Spark version.
Infinitely scalable compute and storage.
Pay for only what you use.
Oracle Confidential – Restricted