LDNWebPerf April 2017 - Andy Still

Serverless is the latest buzzword sweeping the industry but what is it and more importantly, how does it perform?

Transcript

1. Digital Performance
   Serverless
   Performance
   By Andy Still
   

   View Slide

2. Andy Still : Technical Director & Founder, Intechnica
   Web: www.intechnica.co.uk
   Email: andy.still@intechnica.co.uk
   Twitter: @andy_still / @intechnica
   Blog: http://www.internetperformanceexpert.com/
   http://www.intechnica.co.uk/web-performance-warrior.aspx http://oreil.ly/2cJLxZh http://oreil.ly/2ce2ghE
   

   View Slide

3. Digital Performance
   What is Serverless?
   

   View Slide

4. History
   

   View Slide

5. History
   

   View Slide

6. History
   

   View Slide

7. History
   

   View Slide

8. History – Going Virtual
   

   View Slide

9. History – into the cloud
   

   View Slide

10. History – into the cloud
    

    View Slide

11. Now : Serverless
    

    View Slide

12. Digital Performance
    What is AWS Lambda?
    

    View Slide

13. Digital Performance
    What it isn’t
    

    View Slide

14. A lamb
    

    View Slide

15. View Slide

16. A lamborghini
    

    View Slide

17. A C# Lambda Expression
    customers.Where(c => c.City
    == "London");
    

    View Slide

18. Digital Performance
    So, what is it then?
    

    View Slide

19. Digital Performance
    A standalone piece of code that runs in a shared
    infrastructure
    

    View Slide

20. Important Lesson…
    

    View Slide

21. It is not Serverless…
    

    View Slide

22. It does still run on servers…
    

    View Slide

23. You just don’t need to know
    about them
    

    View Slide

24. Better name:
    Function-As-A-Service
    

    View Slide

25. A standalone piece of code that
    runs in a shared infrastructure
    Code is managed and
    deployed as individual
    functions
    Triggered by external events
    

    View Slide

26. A standalone piece of code that
    runs in a shared infrastructure
    You are only responsible for
    your code – the environment
    is outside your control
    No control over compute
    resources used
    You specify the amount of
    memory you want assigned
    to your function
    

    View Slide

27. Multiple Supported Languages
    Can develop in 4 supported
    languages
    Java
    Python
    Node
    C#
    

    View Slide

28. Digital Performance
    Benefits
    

    View Slide

29. Benefits
    No servers to manage
    Code scales to your requirements
    Can execute many times in
    parallel
    Only pay for execution time
    Tracked down to nearest 100ms
    Isolation of functionality
    

    View Slide

30. Digital Performance
    How does it work?
    

    View Slide

31. View Slide

32. Digital Performance
    How can it be triggered?
    

    View Slide

33. Triggers
    Other AWS services
    S3
    Dynamo DB
    Kinesis
    SNS
    Cloudwatch
    External requests
    Via AWS API Gateway
    E.g. triggered from website
    Scheduled
    Proactive or reactive
    

    View Slide

34. Digital Performance
    What’s it for?
    

    View Slide

35. Real time file processing
    

    View Slide

36. Real time Data Processing
    

    View Slide

37. Data Transformation
    

    View Slide

38. Deployment tasks
     Plumbing for a deployment pipeline
     Platform transforms
     Backups
     etc
    

    View Slide

39. Backends
    

    View Slide

40. Digital Performance
    Performance – is it ready for prime time?
    

    View Slide

41. Serverless does not mean no
    servers… Just you don’t have
    to worry about servers
    

    View Slide

42. Whilst you can simplify the
    management of servers and
    environments…
    

    View Slide

43. The requirement to understand
    and analyse performance data
    is as important as ever
    

    View Slide

44. Digital Performance
    Performance Challenges
    

    View Slide

45. Standard server performance
    metrics are no longer available.
    Only stats available relate to
    external execution times
    

    View Slide

46. All functions have to be
    written to be idempotent
    - no cached data or files
    - no connection pooling
    

    View Slide

47. Digital Performance
    Overheads
    

    View Slide

48. Understand other overheads
    as well as the server side
    request…
    Look at RUM / Synthetic data
    

    View Slide

49. Non-server side overhead
    https://blog.newrelic.com/2017/01/11/aws-lambda-cold-start-optimization/
    

    View Slide

50. Digital Performance
    Cold Start
    

    View Slide

51. Functions have to create
    environment on first use
    Known as a Cold Start
    

    View Slide

52. Environment has to be re-
    created after period of no
    activity or if underlying
    infrastructure needs to expand
    under load.
    

    View Slide

53. After approx 5 mins of
    inactivity or 2-4 hours
    regardless of activity
    

    View Slide

54. Performance of regular v Intermittent Functions
    https://blog.newrelic.com/2017/01/11/aws-lambda-cold-start-optimization/
    

    View Slide

55. Each cold start Lambda
    execution adds significant
    overhead
    

    View Slide

56. Cold start overhead varies by
    language
    

    View Slide

57. Cold start overhead can be
    mitigated by minimising
    amount of packages included in
    the build
    

    View Slide

58. Easily mitigated with
    scheduled task
    

    View Slide

59. Digital Performance
    Serverless Performance and Economics
    

    View Slide

60. Serverless is a perfect
    economic reflection of good
    performance
    

    View Slide

61. Serverless is a perfect
    economic reflection of good
    performance
    Dan North
    

    View Slide

62. Serverless Cost Model
    • Number of Executions
    • Memory allocated
    • Execution Time
    3 Elements
    

    View Slide

63. Execution is charged in 100ms
    blocks – optimise performance
    to meet these block limits
    

    View Slide

64. As always… evaluate cost
    savings against cost of
    optimisation
    

    View Slide

65. Digital Performance
    Optimising Memory size
    

    View Slide

66. Memory is the only dial
    available – adjusting memory
    also adjusts the CPU and other
    server settings
    

    View Slide

67. Example: Low memory, High CPU task
    Task:
    In local testing, uses <50mb of memory
    https://serverless.zone/my-accidental-3-5x-speed-increase-of-aws-lambda-functions-6d95351197f3#.q18h6eg9k
    

    View Slide

68. Example: Low memory, High CPU task
    REPORT Duration: 2120.67 ms Billed Duration: 2200 ms Memory Size: 128 MB
    REPORT Duration: 2778.51 ms Billed Duration: 2800 ms Memory Size: 128 MB
    REPORT Duration: 2185.11 ms Billed Duration: 2200 ms Memory Size: 128 MB
    REPORT Duration: 1883.18 ms Billed Duration: 1900 ms Memory Size: 128 MB
    REPORT Duration: 353.52 ms Billed Duration: 400 ms Memory Size: 1536 MB
    REPORT Duration: 340.44 ms Billed Duration: 400 ms Memory Size: 1536 MB
    REPORT Duration: 356.91 ms Billed Duration: 400 ms Memory Size: 1536 MB
    REPORT Duration: 280.49 ms Billed Duration: 300 ms Memory Size: 1536 MB
    REPORT Duration: 336.75 ms Billed Duration: 400 ms Memory Size: 1536 MB
    Before:
    After:
    Increasing memory improved performance 7x
    

    View Slide

69. Digital Performance
    Consistency of Performance
    

    View Slide

70. Lambda functions deliver
    inconsistent performance
    https://blog.symphonia.io/the-occasional-chaos-of-aws-lambda-runtime-performance-880773620a7e#.s1vd0mr6f
    

    View Slide

71. Regular executions at varying memory size
    infrastructure
    

    View Slide

72. Sorted into response time order
    

    View Slide

73. What does this mean?
    Documented performance levels are
    WORST CASE SCENARIO
    

    View Slide

74. What does this mean?
    Performance scales *at least*
    proportionally to memory settings
    

    View Slide

75. What does this mean?
    Higher memory settings result in higher
    levels of consistency
    

    View Slide

76. What does this mean?
    Benchmarking requires long term data
    capture to be sure of reliable
    interpretation
    

    View Slide

77. Digital Performance
    Platform Comparison
    

    View Slide

78. Response Time Under Load
    Azure Functions
    Google Cloud
    AWS Lambda
    http://cloudacademy.com/blog/microsoft-azure-functions-vs-google-cloud-functions-fight-for-serverless-cloud-domination-continues/
    

    View Slide

79. Digital Performance
    Questions
    

    View Slide

80. Andy Still : Technical Director & Founder, Intechnica
    Web: www.intechnica.co.uk
    Email: andy.still@intechnica.co.uk
    Twitter: @andy_still / @intechnica
    Blog: http://www.internetperformanceexpert.com/
    http://www.intechnica.co.uk/web-performance-warrior.aspx http://oreil.ly/2cJLxZh http://oreil.ly/2ce2ghE
    

    View Slide