Trends The Good, The Bad, The Ugly

Maria Colgan Master Product Manager, November, 2019

1 © 2019 Oracle Safe harbor statement

The following is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions.

The development, release, timing, and pricing of any features or functionality described for Oracle’s products may change and remains at the sole discretion of Oracle Corporation.

2 © 2019 Oracle Machine Learning in IoT

Automation Hyperscale Technology Create Database

Cloud Innovations Trends

{{ } } Some Trends are GOOD, JSON Some are BAD, Microservices Some are UGLY REST Map Reduce Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation - Cloud

Computing and Information Services delivered on the Internet Instead of built by customers in their data centers

6 Database Trend - Cloud

Databases running on Cloud Infrastructure Complete automation of infrastructure for databases Automation of Database Lifecycle – create, backup, etc.

7 Database Trend - Cloud Databases

Databases running on Cloud Infrastructure Complete automation of infrastructure for databases Automation of YourDatabase Vote: Lifecycle Cloud – create, Database backup, etc. Good or Bad

8 Database Trend - Cloud Databases

Databases running on Cloud Infrastructure Complete automation of infrastructure for databases Automation of Database LifecycleWe think!– create, backup, etc.

9 Oracle Makes it Better - Optimized Cloud Infrastructure for Databases

Oracle takes Cloud Database to the next level with optimized Cloud Infrastructure for database All Exadata optimizations in the cloud Extreme scalability, availability, security, and cost-effectiveness Bring your existing databases to cloud without risk to quality of service

10 Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation - Cloud Delivered On-Prem

Some customers cannot move to public cloud Regulations, data sovereignty, corporate policies, latency, etc.

Cloud Vendors are bringing Cloud Architecture into the customer’s Data Center at Customer Amazon Outpost Azure Stack

12 Database Trend - Cloud Database On-Prem

Easy migration of databases to cloud model as first step No need to move huge estate of existing applications Enables low latency data services to applications

13 Database Trend - Cloud Database On-Prem

Easy migration of databases to cloud model as first step No need to move huge estate of existing applications Your Vote: Cloud DB On-Prem Enables low latency dataGood services or to Bad?applications Allows database cloud deployments with existing on-premises security controls

14 Database Trend - Cloud Database On-Prem

Easy migration of databases to cloud model as first step No need to move huge estate of existing applications We think! Enables low latency data services to applications Allows database cloud deployments with existing on-premises security controls

15 Oracle Makes it Better - Exadata Cloud at Customer

Oracle takes it to the next level by making deployment and migration super easy Just move databases to Exadata Cloud at Customer Same Oracle Database functionality, interfaces, APIs as Exadata in Public Cloud Same pay-per-use subscription model as Public Cloud

16 Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation – Hyperscale Computing

Demand for Hyperscale computing driven by internet companies who need to run giant distributed sites

18 Database Trend - NoSQL Databases

NoSQL Databases were intended to be: More scalable due to sharded architecture Scale by creating many databases (shards) Choose shard for data by primary key (e.g. user id) More available since every change is atomic No transactions

19 Database Trend - NoSQL Databases

NoSQL Databases were intended to be: More scalable dueYour to sharded Vote: architectureNoSQL Database Scale by creating many databasesGood (shards) or Bad? Choose shard by primary key (e.g. user id) More available since every change is atomic No transactions

20 Database Trend - NoSQL Databases We think Sharding for Hyperscale NoSQL Databases were intended to be: More scalable due to sharded architecture Scale by creating many databases (shards) Choose shardHowever, by primary key limitations (e.g. user id) of NoSQL More available(nosince ACID, every key change-value is atomic only, etc.) are No transactions

21 Oracle Makes it Better –Native Sharding

Native Sharding in Oracle provides benefits of sharding with all the benefits of a mature SQL Database Ability to create up to 1000 Shards Linear scalability of data and users One Massive DB Online addition and reorganization of shards to Many Small DBs Routing of SQL based on shard key, Cross Shard Queries Native support for data sovereignty by region or country

Shard #1 Shard #2 Shard #3 Americas EMEA Asia

22 Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices { } 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation - JSON

Developers like flexible data model Don’t want to declare fixed set of tables, {"id":1, columns, etc. (no schema) "name":"Century 16", "location":{"street":"Main St", "city":"Redwood", Want to store and access data "zipCode":"94607", using the format used by application "state":"CA", "phoneNumber":null }, Developers don’t want complexity of } managing a

24 Database Trend - Document Databases

Nonrelational database designed to store and query data as JSON documents Simple for developers to create and manage

25 Database Trend - Document Databases

Nonrelational database designed to store and query data as JSONYour documents Vote: Document Database Simple for developers toGood create or and Bad? manage

26 Database Trend - Document Databases

Nonrelational Wedatabase think designed JSON datato store type and isquery data as JSON documents Simple for developers to create and manage But Document Databases are

27 Oracle Makes it Better – JSON in RDBMS

Relational databases now have native support for JSON and XML Combine all the benefits of relational with benefits of documents Oracle stores JSON documents in table columns, with native SQL support SELECT c.json_column.address.city { } FROM customers c; JSON supported by Analytics, Encryption, In-Memory, RAC, Parallel SQL, … Can index and edit any JSON element without replacing the entire document JSON_DataGuide discovers structure of JSON documents and created virtual columns Autonomous Database solves management complexity of relational Databases

28 Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation – Parallel Map Reduce

Framework to query and process large volumes of data

30 Database Trend – Hadoop / Spark

Database-like platform for high-scale, batch and analytics Low cost storage - Data Lake on HDFS and Object Store

Object Store

DATA LAKE

31 Database Trend – Hadoop / Spark

Database-like platform for high-scale, batch and analytics Low cost storageYour - HDFS Vote: and ObjectsHadoop Stores / Spark Good or Bad?

32 Database Trend – Hadoop / Spark

Database-like platform for high-scale, batch and analytics Low cost storage We- HDFS think and HadoopObjects Stores is Ugly

Promised a Data Lake, got a Data Swamp

33 Database Trend – Hadoop / Spark

Map Reduce is super hard to program, so being replaced SQL-like query languages (faux SQL) If you are mostly running SQL then better to use a real SQL DB: More powerful SQL, better joins, better optimizers, more sucure, etc.

34 Oracle Makes it Better – CloudSQL on Object Stores

Expand database to include external data in ORACLE HDFS or object store data lakes DATBASE Sophisticated Oracle scalable SQL over Oracle DW and object store Oracle Object Store, AWS S3, Azure Blobs Supports all popular file formats Including CSV, JSON, Parquet and Avro Ability to call out to Machine Learning libraries Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation – Microservices

In microservices, applications are written as independent services Each development team can rapidly develop and evolve their microservice independently

37 Database Trend – Separate Database for each Service

Separate database for each service Select a database specifically designed for each service use case

38 Database Trend – Separate Databases for each Service

Separate database for each service Select databaseYour that is Vote: specifically Are designedmicroservices for service use case with separate databases good or bad?

39 Database Trend – Separate Databases for each Service

Separate databaseWe for think each serviceMicroservices are Select database that is specifically designed for service use case

We think Separate Databases for Microservices are

40 Database Trend – Separate Databases for each Service

Integration of separate micro-service databases creates “macro-level” complexity You need to administer, secure, tune and patch each database separately

41 Oracle Makes it Better – Convergeable Microservices

Convergeable Microservice Databases provide independence without integration complexity Microservices are developed as if databases are separate Developers focus on application logic rather than database integration

30 Oracle Makes it Better – Convergeable Microservice

Databases can be flexibly combined or separated

Combining is enabled by the ability to converge many databases, data types, and workloads into one converged container database

Container Database

43 Oracle Makes it Better – Convergeable Microservice

Separation of databases is enabled by using Pluggable Databases that can be dynamically moved between physical container databases

Container Database Container Database Container DB

44 Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation – Machine Learning

Machine Learning is able to solve analytics problems and identify patterns that were previously very difficult or impossible Anticipate customer behavior Detect fraud Personalize shopping or advertisements

Note: not discussing ML for unstructured data here Image processing, speech recognition, etc.

34 Database Trend – Specialized ML Data Platform

Data moved from databases to specialized analytic platforms to run Machine Learning SAS, SPARK, R, etc.

35 Database Trend – Specialized ML Data Platform

Data moved from databases to specialized platforms to run Machine LearningYour Vote: Are specialized ML SAS, SPARK, R, etc. Data Platforms good or bad?

35 Database Trend – Specialized ML Data Platform

Data movedWe from think databases specialized to specialized ML platforms Data to run Machine Learning Platforms are SAS, SPARK, R, etc.

Better to bring algorithms to data instead of data to algorithms

35 Oracle Makes it Better – In-Database ML

Oracle natively includes dozens of ML algorithms Perform parallel ML directly in Data Warehouse or OLTP for fast model building and real time scoring of new data Keep data secure by avoiding copy contagion – copies of data in analytics systems are a common source of data breaches Obstacles to adoption of ML in database have been removed Ability to call out to latest R and Python algorithms Accessible via SQL, REST API, Oracle Data Miner GUI, Oracle R Enterprise

50 Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation – Blockchain

Blockchain is a growing list of records that are cryptographically linked to resist unauthorized changes and provide a verified history of changes

Requires consensus among participants to make a change

Originally designed to serve as public transaction ledger for bitcoin

52 Database Trend – Distributed Blockchain Platforms

Specialized platforms that hold blockchain protected data and run blockchain based applications Making a change requires consensus of participants

53 Database Trend – Distributed Blockchain Platforms

Specialized platforms that hold blockchain protected data and run blockchain basedYour applications Vote: Are Distributed Making a change requiresBlockchain consensus Platforms of participants good or bad?

54 Database Trend – Distributed Blockchain Platforms

SpecializedWe platforms think that Blockchain hold blockchain ledger protected is good data and run blockchain based applications Making a change requires consensus of participants

But Distributed Blockchain is too complex for most enterprise needs

55 Oracle Makes it Better – Native Blockchain Tables

• Coming in Database 20c: • CREATE Blockchain Table Simple – specialized table that is insert-only Trade_Ledger; • Rows are cryptographically chained • Chain is verifiable by participants Blockchain

• Blockchain Table can participate in transactions and queries with other tables • Orders of magnitude easier than distributed block chains

56 Oracle Makes it Better – Native Blockchain Tables

No Trust Use when participants don’t trust each other, but partially trust a neutral service provider Examples: Escrow company, SaaS service, Trade Exchange Participants get application transparent protection from fraud by other participants Trust But Verify Participants can sign and verify data to detect provider fraud (trust but verify) Provider Satisfies majority of enterprise blockchain use cases today

57 Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation – Internet of Things

All assets and devices are connected and share information Phones, tablets, smart meters, cars, fitness devices Generates vast amounts of data Rapid analysis of IoT data provides a competitive edge

59 Database Trend – In-Memory Database for IoT

In-Memory database to capture high-volume IoT data Memory is an order of magnitude faster than disk making it ideal for fast data ingest OK to lose some data

60 Database Trend – In-Memory Database for IoT

In-Memory database to capture high-volume IoT data Memory is an orderYour of magnitudeVote: Are faster In-Memory than disk making it ideal for fast data ingestDatabases for IoT good or bad? OK to lose some data

61 Database Trend – In-Memory Database for IoT

In-Memory databaseWe think to capture In-Memory high-volume ingest IoT data is Memory is an order of magnitude faster than disk making it ideal for fast data ingest OK to lose some data But pure In-Memory IoT limits capacity and increases cost

62 Oracle Makes it Better – IoT Data Streaming

Example: Insert Temperature Sensor Readings IoT New in-memory insert algorithm Insert: TEMPERATURE <6:05AM, 55o > SENSOR Declare table MEMOPTIMIZE FOR WRITE

In-Memory SQL Insert performs low-latency insert Ingest Buffer into in-memory buffer Time Temp In-Memory 05:50 52o Bulk insert done asynchronously in Append 05:55 54o background to low cost storage 06:00 54o Ultra-fast 06:05 55o 25 million inserts per second on two DRAINER PROCESSES PERIODICALLY socket server DRAIN BUFFER TO DISK Speed of In-Memory, cost of Storage TEMPERTURE READINGS Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation – REST as Messaging API

Text-based stateless protocol with a simple request/response model Request Small set of operations:

GET, POST, PUT, DELETE Response Makes communication simple

65 Database Trend - REST as Data API

REST used to invoke common database operations select, insert, etc. REST used to invoke SQL or database stored procedures Makes database access simple

66 Database Trend - REST as Data API

REST used for common database operations – select, insert, etc. Your Vote: Is REST as Data REST used to invoke SQL or database stored procedures Interface good or bad? Makes database access simple

67 Database Trend - REST as Data API

REST used for common database operations – select, insert, etc. REST used to invokeWe SQL think or database REST storedas Data procedures Makes database access simpleInterface is good

68 Oracle Makes it Better – Oracle REST Data Services

URI Map & Bind SQL

JSON Transform to JSON SQL Result Set

HTTPS Client REST Data Services Oracle Database

Automatic REST Data Services creates REST endpoints that map to SQL or PS/SQL e.g. SELECT ….FROM customers , orders WHERE …

Transforms SQL results into JSON or other formats (CSV, etc.)

69 Technology Innovations

1 Cloud 2 Cloud Delivered On-Premises 3 Hyperscale Computing 4 JSON as Standard Data Format 5 Parallel Map Reduce 6 Microservices 7 Machine Learning for Analytics 8 Blockchain 9 Internet of Things 10 REST as Messaging API 11 Self-Driving Vehicles Technology Innovation – Self Driving Vehicle

Self-driving vehicles navigate themselves through a complex and changing environment

71 Database Trend - Self-Driving Database

Self-driving databases manage and optimize themselves through complex and changing workloads

72 Database Trend - Self-Driving Database

Self-driving databases manage and optimize themselves through complex and changingYour Vote: workloads Can self-driving Vehicles and Databases handle the complexity of the real-world

73 Database Trend - Self-Driving Database We think self-driving vehicles will be Self-drivingchallenged databases manageby complexity and optimize of themselves real-world through complex and changing workloads

ML plus decades of management automation enable self-driving databases

74 Oracle Makes it Better – Mission-Critical Autonomous Database

Self-Driving Self-Securing Self-Repairing • Scale-out database with • Automatically applies • Recovers automatically fault-tolerance and DR security updates online from any failure • Runs on enterprise- • Secure configuration • 99.995% uptime proven Exadata platform with full database including maintenance encryption • Full compatibility with • Elastically scales existing enterprise • Sensitive data hidden compute or storage as databases from Oracle or customer needed admins Machine Summary Learning Cloud in IoT Data Center Technology innovations create new database trends Hyperscale Automation Some database trends are Technology good, some bad, some ugly

Cloud Innovative data technology is InnovationsBlockchain often tied to limited single- purpose databases {{ } } Oracle is adopting the good JSON Microservices parts of innovative data tech in its leading enterprise database REST Map Reduce