High Performance Architecture Patterns

Transcript

1. High Performance
   Architecture Patterns
   Celso Crivelaro
   @celsocrivelaro
   

   View Slide

2. $ whoami > Celso Crivelaro
   - Engineer and MSc in Computer Engineering
   - Dev Ruby / Python / Java
   @celsocrivelaro
   http://crivelaro.me
   

   View Slide

3. View Slide

4. View Slide

5. SMTP Locaweb
   - Transaction email platform
   - Fast and high quality email sending
   - Live reports
   - Integration by SMTP Protocol and HTTP API
   - 2,5 Millions of messages per day
   - + 2 years
   

   View Slide

6. View Slide

7. Platforms
   

   View Slide

8. View Slide

9. View Slide

10. View Slide

11. View Slide

12. High Availability
    

    View Slide

13. Resource Consumption
    

    View Slide

14. View Slide

15. Disk
    Memory
    Band-
    width
    Processor
    time
    Logs
    GC
    Open
    Connections
    Deadlocks
    Stop all disk activities
    Swap
    Slowness
    Services stop
    responding
    Queues Amount
    Workers
    Slowness
    

    View Slide

16. Integrations
    

    View Slide

17. Integrations always
    fails!
    

    View Slide

18. Integration Fails always
    Internal
    External
    DB API Storage Network
    API ISP Hop
    Cloud
    

    View Slide

19. Map hidden
    integrations
    E.g. DNS
    

    View Slide

20. Fail Fast
    

    View Slide

21. Failing Fast reduces
    transient errors
    

    View Slide

22. Timeouts
    

    View Slide

23. Never trust in
    default timeouts
    

    View Slide

24. Threads Timeouts to get rid of deadlocls
    Internal Little timeout
    External Depends: The least possible
    

    View Slide

25. Retry
    

    View Slide

26. 1 2 3 5 8 n
    Fibonacci Proportion
    ...
    

    View Slide

27. Isolate Critical
    Components
    

    View Slide

28. Panel / API
    Painel Web Internal API - HTTP
    Resources
    Connection
    Pool
    App Server
    Business
    Rules
    Integrations
    

    View Slide

29. App Internal API
    App Panel
    Painel Web API - HTTP
    Resources
    Conn.
    Pool
    App
    Server
    Business
    Rules
    Integrations
    Resources
    Conn.
    Pool
    App
    Server
    Business
    Rules
    Integrations
    

    View Slide

30. Less Components
    ==
    Less Problems
    

    View Slide

31. View Slide

32. Redundancy
    

    View Slide

33. Health Check
    

    View Slide

34. View Slide

35. View Slide

36. Scalability
    

    View Slide

37. CAP Theorem
    

    View Slide

38. Availability
    Consistency
    Tolerance
    to network
    partition
    

    View Slide

39. Availability
    Consistency
    Partition
    Tolerance
    to network
    partition
    whether system is
    available or not
    

    View Slide

40. Availability
    Consistency
    Tolerance
    to network
    partition
    Same data in all nodes
    

    View Slide

41. Availability
    Consistency
    Tolerance
    to network
    partition
    Tolerante when a network
    fails
    

    View Slide

42. Availability
    Consistency
    Partition
    Tolerance
    Postgres
    

    View Slide

43. Availability
    Consistency
    Partition
    Tolerance
    DNS
    

    View Slide

44. Availability
    Consistency
    Partition
    Tolerance
    Protocols
    

    View Slide

45. Evaluate which one is
    acceptable to your
    platform
    

    View Slide

46. Pool of Resources
    

    View Slide

47. App
    Connection
    Resource
    Connection
    Connection
    Connection
    Connection
    

    View Slide

48. Controlled Timeouts
    Controlled Resource Consumption
    Avoid opening/closing connections
    overhead
    

    View Slide

49. Cache
    

    View Slide

50. Make your data closer
    to your app
    

    View Slide

51. When use cache
    App API
    DB
    

    View Slide

52. Default Cache Strategy
    App
    Cache
    Source
    1
    2
    3
    

    View Slide

53. Hot Cache Strategy
    App Cache Source
    2 1
    

    View Slide

54. Load Balacing
    

    View Slide

55. LB
    App
    App
    App
    App
    

    View Slide

56. Load Balance Strategy
    matters!
    

    View Slide

57. Decoupled
    Architecture
    

    View Slide

58. Slow Processing
    Fast processing
    Asynchronous Processing
    Authentication
    Message
    Transformation
    MTA Providers
    User
    Message
    Persistence
    Consolidation
    

    View Slide

59. Background jobs
    Messaging frameworks
    

    View Slide

60. How to discover
    bottlenecks?
    

    View Slide

61. Test your limits...
    Load Test
    Stress Test
    Peak Test
    

    View Slide

62. Telemetry
    

    View Slide

63. View Slide

64. Grafana + InfluxDB
    

    View Slide

65. Other scalabity
    patterns
    

    View Slide

66. Indexes
    

    View Slide

67. Sharding
    

    View Slide

68. References
    

    View Slide

69. View Slide

70. View Slide

71. View Slide

72. http://use-the-index-luke.com
    

    View Slide

73. View Slide

74. Thank you!
    Celso Crivelaro
    @celsocrivelaro
    celsocrivelaro@gmail.com
    

    View Slide