The document discusses using event streams and read models to address restrictions of the classical approach to building business applications. It describes how CQRS and event sourcing can provide benefits like high alignment with business needs and decoupling of read models. The NuClear River framework is presented as an implementation of these concepts that uses an OData API and pipeline-based processing to provide a performant solution while controlling data consistency.

2. Building Read Models
using event streams
https://github.com/denisivan0v
@denisivanov

3. About me

4. About me
Applied math and computer science

5. About me

6. Business Domain
Sales

7. Business Domain
- Sales staff with different
responsibilities

8. Business Domain
- Sales staff with different
responsibilities
- Many business processes and
workflows

9. Business Domain
- Sales staff with different
responsibilities
- Many business processes and
workflows
- Consistency checks (validations)

10. Business Domain
- Sales staff with different
responsibilities
- Many business processes and
workflows
- Consistency checks (validations)
- Reports

11. Classical approach
User Interface
Business Logic
Database

12. Classical approach
User Interface
Business Logic
Database
SQL Server

13. Classical approach
User Interface
Business Logic
Database
SQL Server
ORM

14. Classical approach
User Interface
Business Logic
Database
SQL Server
Repositories
Services
ORM

15. Classical approach
User Interface
Business Logic
Database
SQL Server
Repositories
Services
ASP.NET MVC
+ WCF + JS
ORM

16. Restrictions
- Queries on data model cause loss
of meaning

17. Restrictions
- Queries on data model cause loss
of meaning
- Balance between reads and writes

18. Restrictions
- Queries on data model cause loss
of meaning
- Balance between reads and writes
- Changes in the data model

19. Restrictions
- A lot of parameterized querying
criteria

20. Restrictions
- A lot of parameterized querying
criteria
- Generic query API

21. Restrictions
- A lot of parameterized querying
criteria
- Generic query API
- Querying data from many systems
or derived data

22. Another approach
Command and Query
Responsibility
Segregation

23. CQRS approach
User Interface
Business Logic
Database

24. CQRS approach
User Interface
Business Logic
Database
Thin Read Layer

25. CQRS approach
Bounded context

26. CQRS approach
User Interface
Business Logic
Database
Thin Read Layer

27. CQRS approach
User Interface
Business Logic
Database
Thin Read Layer
Database

28. CQRS approach
User Interface
Business Logic
Database
Thin Read Layer
Database

29. CQRS approach
User Interface
Domain Model
Event Store
Thin Read Layer
Database

30. CQRS approach
Eventual consistency

31. CQRS approach
User Interface
Business Logic
Database
Thin Read Layer
Database
Read Model

32. CQRS approach
User Interface
Write Model Read Model

33. CQRS approach
User Interface
Write Model Read Model
Read Model
Read Model

34. Moving to Read Model
User Interface
Business Logic
Database

35. Moving to Read Model
User Interface
Business Logic
Database
Events

36. Moving to Read Model
User Interface
Business Logic
Database
Events
Observer

37. Moving to Read Model
User Interface
Business Logic
Database
Events
Observer
Database

38. Moving to Read Model
User Interface
Business Logic
Database
Events
Observer
Database

39. Moving to Read Model
User Interface
Business Logic
Database
Events
Querying
API
Observer
Database

40. Moving to Read Model
NuClear River
https://github.com/2gis/nuclear-river

41. NuClear River
User Interface
Business Logic
Database
Events
Querying
API
Observer
Database

42. NuClear River
User Interface
Business Logic
Database
Querying
API

43. NuClear River. Querying
- REST API based on OData
protocol

44. NuClear River. Querying
- REST API based on OData
protocol
- WebAPI + OData libs + EF stack

45. NuClear River. Querying
- REST API based on OData
protocol
- WebAPI + OData libs + EF stack
- Driven by Read Model description
(DSL)

46. NuClear River. Querying
Demo

47. NuClear River. Observer
- Pipeline-based

48. NuClear River. Observer
- Pipeline-based
- Separated facts storage

49. NuClear River. Observer
- Pipeline-based
- Separated facts storage
- Controls invariants when building
aggregates in Read Model

50. NuClear River. Observer
- Pipeline-based
- Separated facts storage
- Controls invariants when building
aggregates in Read Model
- Also driven by descriptions (DSL)

51. NuClear River. Telemetry
- Uses ELK
(Elasticsearch+Logstash+Kibana)
stack

52. NuClear River. Telemetry
- Uses ELK
(Elasticsearch+Logstash+Kibana)
stack
- Provides an ability to publish any
tracing information

53. NuClear River. Querying
Demo

54. Profits
- High alignment with business

55. Profits
- High alignment with business
- Proper level of Read Model
decoupling

56. Profits
- High alignment with business
- Proper level of Read Model
decoupling
- Open standard API

57. Profits
- High alignment with business
- Proper level of Read Model
decoupling
- Open standard API
- Performance

58. Summary
- Classical approach works, but not
always

59. Summary
- Classical approach works, but not
always
- Start from business case,
determine bounded context

60. Summary
- Classical approach works, but not
always
- Start from business case,
determine bounded context
- Choose the right pattern

61. Summary
- Classical approach works, but not
always
- Start from business case,
determine bounded context
- Choose the right pattern
- Use NuClear River to save costs

62. Thank you!
Denis Ivanov
https:// github.com/denisivan0v
@denisivanov

63. NuClear River
https://github.com/2gis/nuclear-river
https://2gis.gitbooks.io/nuclear-
river/content/en/index.html