Scale Cube ACID vs BASE CAP Theorem CQRS Event Sourcing klunge.io

1. Open Source Event Sourcing Framework
Derya SEZEN
Cloud/DevOps Consultant @kloia
derya_sezen

2. Problem?
Scaling

3. Bottlenecks?
In-memory
limits
Storage
limits
NetworkGeoIP
HW
I/O OS
Unicast
vs
Anycast

4. Solution?

5. How you will scale?

6. How you will scale?

7. How you will scale?

8. Scale Cube

9. Why not invest on Scale Cube?

10. Scale Cube
● X-axis: This is usually referred as
scale-out and scale-in
● Y-axis: Any idea? Any example?

12. Is your architecture ready to be decoupled?

13. Is your business/data/load is growing?

14. “Global mobile users to exceed 5.5B by
2022” -- Forrester
Smartphone New Generation 5G

15. Is your architecture
ready to be
decoupled?

16. Is your
organisation
ready to be
decoupled?

17. Is 2 years
sufficient for
such major
change?

18. Does your
business
convinced to
invest on that
non-functional
change?

19. Are Microservices the solution to
scaling?

20. Microservices is not an
architecture!

21. ACID vs BASE
Atomic: All actions in Xact happen, or none
happen
Consistency: If each Xact is consistent, and the
DB starts consistent, it ends up consistent
Isolation: Execution of one Xact is isolated from
that of other Xacts.
Durability: If a Xact commits, its effects persists.
Basic Availability: Read/Write functions
available w/ no consistency guarantee
Soft State: After some time, we know the state
Eventual Consistency

22. non-ACID examples?
Twitter likes/timelines
Facebook news
Amazon Cart/Payment

23. CAP Theorem
Consistency Availability
Partition
Tolerance
N/A
CA
CP AP

24. Partition Tolerance (Referring to Scale Cube)

25. Consistency
Stock = 1
Stock = 1
Replication
Checkout
Product Page

26. Inconsistency
Stock = 0
Stock = 1
Replication
Checkout
Product Page

27. Inconsistency
Stock = 0
Stock = 1
Broken link!
Checkout
Product Page

28. Choose Availability or Consistency?
A or C?
Availability Consistency
Continue with
Inconsistency Return Error to User

29. “Availability over consistency is not a
technical choice, it is a business choice.”
-- Anonymous

30. Ask your business?
Which one you prefer?
1- The website goes down
2- Website shows products with Stock=0

31. CAP Theorem
Consistency Availability
Partition
Tolerance
N/A
CA
CP AP
RDBMS
CP AP

32. Majority is CA
Strong Consistency!

33. Upcoming decade challenge
Consistency Availability
Partition
Tolerance
N/A
CA
CP AP
RDBMS
CP AP

34. In case P
Failure probability increases exponentially with
the # of nodes

35. “Anything that can go wrong will go wrong”
-- Murphy's law

36. CQRS
Command–query separation
Read DB Write DB
UI / API

37. CQRS
Command–Query Responsibility Segregation
Ref: https://altkomsoftware.pl/en/blog/cqrs-event-sourcing/

38. Event-Driven ≠ Event-Sourcing

39. Ref: https://www.nginx.com/blog/event-driven-data-management-microservices/
Event-Driven

40. What makes Event-Sourcing
different?

41. Entities & Events
Account 1, +50
Account 1, +150
Account 2, +300
Account 1, -100
Account
Number
Balance Description
1 100 Bank A
2 500 Bank B
….
Table: Account
Events: Account
Account 2, +200

42. CQRS with Event Sourcing
Ref: https://altkomsoftware.pl/en/blog/cqrs-event-sourcing/

43. CQRS ≠ Event-Sourcing

44. CQRS opens the path to Event-Sourcing

45. Event-Sourcing
Recap:
● Every state change recorded as an event
● The event store becomes the principal source of truth
● State = rebuild using events
● Advantages:
○ No additional Audit Logs, Event are audit logs!
○ Highly-scalable!

46. Event-Sourcing
Challenges?

47. Challenges of Event Sourcing
● Operational Consistency
● Configurable Snapshot Policies & Mechanisms
● Standardized Rollback Mechanisms
● Concurrency Resolving
● Topology Design
● Data Migration

48. Why do you need a Framework/Tool for ES?
● Same reason you need framework(s) on your software.
● Decrease the learning-curve

49. What functions do you need from the
frameworks?
● Guarantee Data consistency
● Snapshots: Build snapshots of the working copy
● Rollbacks

50. Without a framework, will you fail?

51. We developed klunge.io to solve those complexities
Contributions are welcome:
https://github.com/kloiasoft/eventapis

52. Thank you!
Derya SEZEN
Cloud/DevOps Consultant @kloia
derya_sezen klunge.io