Domain-Driven Data at That Conference

There are many types of databases and data analysis tools from which to choose today. Should you use a relational database? How about a key-value store? Maybe a document database? Or is a graph database the right fit for your project? What about polyglot persistence? Help!

Applying principles from Domain-Driven Design such as strategic design and bounded contexts, this presentation will help you choose and apply the right data layer for your application's model or models. We will explore traditional relational databases, graph databases, document databases, key/value stores, polyglot persistence, CQRS, event sourcing, and data layers for microservices.

Transcript

1. Bradley Holt, Developer Advocate
   Tuesday, August 9, 2016
   @BradleyHolt
   Domain-Driven Data
   

   View Slide

2. View Slide

3. @BradleyHolt
   

   View Slide

4. View Slide

5. View Slide

6. View Slide

7. Big Data
   Get it?
   

   View Slide

8. IBM Cloud Data Services
   Open for Data
   A comprehensive por.olio of open source data services
   

   View Slide

9. A Brief History of Data
   

   View Slide

10. The Relational Database
    @BradleyHolt
    order
    *order_id
    customer_id
    date
    customer
    line_item
    *customer_id
    email_address
    name
    *order_id
    *item_id
    price
    quantity
    

    View Slide

11. ACID Guarantees
    Relational databases guarantee atomicity, consistency, isolation and durability
    

    View Slide

12. Big Iron
    The ACID guarantees provided by relational databases were (and often still are) critical
    for systems of record
    

    View Slide

13. The World Wide Web
    The introduction of the Web brought a whole new type of application with
    different constraints than systems of record
    @BradleyHolt
    

    View Slide

14. Mobile Apps
    The introduction of mobile apps added to the growing number of systems of engagement
    

    View Slide

15. Changing Constraints
    

    View Slide

16. Always On
    

    View Slide

17. Big Data
    

    View Slide

18. The CAP Theorem
    @BradleyHolt
    Partition
    Tolerance
    Availability
    Consistency
    Consistency Availability
    Partition
    Tolerance
    

    View Slide

19. Horizontal Scaling
    Horizontal scaling is scaling through the addition of commodity hardware
    

    View Slide

20. Eventual Consistency
    Given no new updates, each node in a distributed system will eventually have a
    consistent view of the data
    

    View Slide

21. Enter "Not only SQL" (NoSQL)
    

    View Slide

22. @BradleyHolt
    key-value
    graph
    document
    …more
    

    View Slide

23. @BradleyHolt
    

    View Slide

24. Key-Value Stores
    Opaque data accessed through unique keys
    

    View Slide

25. Document Databases
    A variation on key-value stores with strictly defined values (e.g. JSON objects)
    

    View Slide

26. Graph Databases
    Nodes and properties of nodes connected via edges
    

    View Slide

27. Domain-Driven Design (DDD)
    

    View Slide

28. @BradleyHolt
    

    View Slide

29. Domain-Driven Design
    A collaboration between domain experts and software practitioners
    

    View Slide

30. Complexity is in the Domain
    Complexity is in the domain, not in the technology
    

    View Slide

31. Models as Tools
    Models are tools used to solve problems within the domain
    

    View Slide

32. The Map is not the Territory
    Don't confuse models with reality itself
    

    View Slide

33. Building Blocks of DDD and the
    Life Cycle of a Domain Object
    

    View Slide

34. Entities
    Entities are defined by their identity
    

    View Slide

35. Value Objects
    Value objects encode attributes that describe things
    

    View Slide

36. Aggregates
    Aggregates group related entities to minimize complexity
    

    View Slide

37. Repositories
    A repository provides the illusion of an in-memory data store
    

    View Slide

38. Domain Layer
    Order Aggregate
    @BradleyHolt
    «interface»
    OrderRepository
    + insertOrder(order:Order)
    + updateOrder(order:Order)
    + findOrderById(id:int) : Order
    + recentOrders([limit:int]) : Order[0..*]
    Customer
    …
    Infrastructure Layer
    LineItem
    …
    Order
    - id : int
    - customer : Customer
    - date : Date
    - lineItems : LineItem[1..*]
    + total() : Money
    InMemoryOrderRepository
    + insertOrder(order:Order)
    + updateOrder(order:Order)
    + findOrderById(id:int) : Order
    + recentOrders(limit:int) : Order[0..*]
    RelationalMapperOrderRepository
    + insertOrder(order:Order)
    + updateOrder(order:Order)
    + findOrderById(id:int) : Order
    + recentOrders(limit:int) : Order[0..*]
    

    View Slide

39. Choosing the Right Data Layer
    

    View Slide

40. Data Store
    A repository cannot abstract the constraints of your data store
    

    View Slide

41. Object-Relational Impedance Mismatch
    Object-oriented programming and relational databases use different models
    

    View Slide

42. Eric Evans on NoSQL
    "This is the world of NoSQL to me, that we can choose a tool that fits well with
    the problem we're trying to solve." –Eric Evans (author of Domain-Driven Design)
    @BradleyHolt
    

    View Slide

43. Strategic Design
    

    View Slide

44. Bounded Context
    Bounded contexts allow different domain models to be used within different contexts
    

    View Slide

45. One Data Layer Per Bounded Context
    Each bounded context should have its own data layer, and should not directly access a
    data layer belonging to another bounded context
    

    View Slide

46. Data Systems
    A data layer may be a database, or it can be a data system consisting of multiple databases
    

    View Slide

47. Microservices as Bounded Context
    Represent each bounded context as a microservice or a cluster of microservices
    

    View Slide

48. @BradleyHolt
    Catalog
    Document Database
    Key/Value Store
    Graph Database
    Full Text Search
    Shopping
    Cart
    Document Database
    Key/Value Store
    Orders
    Relational Database
    Big Data Analytics (e.g.
    Apache Spark)
    

    View Slide

49. Alternative Architectures
    

    View Slide

50. Command Query Responsibility Segregation (CQRS)
    Rather than update an entity in place, CQRS provides separate read and write models
    

    View Slide

51. Domain Layer
    @BradleyHolt
    Read Model
    «interface»
    OrderQueryHandler
    + findOrderById(id:int) : OrderDetails
    + recentOrders([limit:int]) : OrderSummary[0..*]
    Write Model
    OrderDetails
    + getId() : int
    + getCustomer() : Customer
    + getDate() : Date
    + getLineItems() : LineItem[1..*]
    + getTotal() : Money
    «interface»
    OrderCommandHandler
    + handle(command:CreateOrder)
    + handle(command:AddLineItem)
    CreateOrder
    - customer : Customer
    - date : Date
    - lineItems : LineItem[1..*]
    OrderSummary
    + getId() : int
    + getDate() : Date
    + getTotal() : Money
    AddLineItem
    - orderId : int
    - lineItem : LineItem
    

    View Slide

52. Event Sourcing
    The current application state is computed from a sequence of events
    

    View Slide

53. IBM Cloud Data Services
    Open for Data
    A comprehensive por.olio of open source data services
    

    View Slide

54. View Slide

55. Image Credits
    §  Open for Data Dome (outside) by Bradley Holt
    §  Open for Data Dome (inside) by Bradley Holt
    §  Brent Spiner (Data from Star Trek: The Next Generation) with
    Zoltar from Big by Bradley Holt, on Twitter
    
    §  database 2 by Tim Morgan, on Flickr
    §  Hard Disk by Jeff Kubina, on Flickr
    §  IBM 360 Announcement center by Robert Nix, on Flickr
    
    §  Dialing Up Web History by Mike Licht, on Flickr
    
    §  Instagram and other Social Media Apps by Jason Howie, on Flickr
    
    §  Dynamo, un siècle de lumière et de mouvement dans l'art, 1913 – 2013
    - Galeries nationales du Grand Palais - Paris - 10 avril au 22 juillet 2013
    by Yann Caradec, on Flickr §  World travel and communications recorded on Twitter by Eric Fischer,
    on Flickr
    §  Server grill with blue light by David Precious, on Flickr
    
    §  Spider Web by Alden Chadwick, on Flickr
    §  database by Tim Morgan, on Flickr
    §  Keys for the Stanley Hotel by Mike Silva, on Flickr
    
    §  paper by malik, on Flickr
    §  Edinburgh Road Network analysis by Steven Kay, on Flickr
    
    @BradleyHolt
    

    View Slide

56. Image Credits (cont'd)
    §  IMG_2619 by Jason Pelletier, on Flickr
    §  Sounds_of_Complexity11.jpg by Enzo Varriale, on Flickr
    
    §  model by MaZzuk, on Flickr
    §  taking the subway to find the rents by Eli Duke, on Flickr
    
    §  DSC_3407 by Mad House Photography, on Flickr
    
    §  red numbers by DaveBleasdale, on Flickr
    §  Social graph by Dmitry Grigoriev, on Flickr
    §  Catalog. by Adam Mayer, on Flickr
    §  Lina Bo Bardi, SESC Pompéia by paulisson miura, on Flickr
    
    §  Financial District Classical Building Reflection Distortion, San Francisco,
    California, USA by Wonderlane, on Flickr
    §  Eric Evans by Oliver Gierke, on Flickr
    §  rectangles by Dean Hochman, on Flickr
    §  Hexagons by Henry Burrows, on Flickr
    §  Rooted by Anna Levinzon, on Flickr
    §  Rainforest Biome by BMiz, on Flickr
    §  rectangles-10 by Karen Cropper, on Flickr
    §  Rusty Chain by veggiefrog, on Flickr
    @BradleyHolt
    

    View Slide

57. Questions?
    @BradleyHolt
    

    View Slide

58. View Slide