Discovering & solving performance issues in PHP-apps

I want to lay out how to effectively anticipate, identify and mitigate performance issues in a common PHP application. I want to discuss how to approach performance optimizations including micro-optimizations, some tools available to a developer for investigating and monitoring performance issues as well as some ways to improve performance, including a quick introduction to both application and HTTP-caching.

Transcript

1. DISCOVERING & SOLVING
   PERFORMANCE ISSUES
   

   View Slide

2. I will give you…
   a quick tour of some tools I use
   an opinionated guide of practices
   an overview of typical problems/solutions
   

   View Slide

3. DENIS
   BRUMANN
   denis.brumann@sensiolabs.de
   @dbrumann
   

   View Slide

4. BEFORE YOU START…
   FIGURE OUT YOUR PROBLEM
   

   View Slide

5. View Slide

6. It can be used to simulate a heavy load on
   a server, group of servers, network or
   object to test its strength or to analyze
   overall performance under different load
   types.
   https://jmeter.apache.org/
   

   View Slide

7. LOAD TEST

   Verify application behavior under normal and peak load conditions.
   STRESS TEST 

   Determine or validate an application’s behavior when pushed

   beyond normal or peak load conditions.
   CAPACITY TEST

   Determine how many users and/or transactions a given system

   will support and still meet performance goals.
   

   View Slide

8. View Slide

9. View Slide

10. View Slide

11. View Slide

12. WEB SERVER 

    access & error logs
    DATABASE 

    size, throughput, connections, slow queries
    CACHE

    hits, misses, size
    SYSTEM

    CPU & memory usage, disk I/O, network
    

    View Slide

13. WHAT TO DO?
    slowly increase sample size
    monitor changes & spot anomalies
    make one change at a time

    based on a verifiable assumption
    repeat
    

    View Slide

14. Search for causes not symptoms
    Do not make changes without cause
    Change as few things as possible
    Know when to stop
    

    View Slide

15. You have rock solid & kickass monitoring

    for your production system!
    Tests & monitoring can be reused & adapted

    for other scenarios
    

    View Slide

16. Performance Testing Guidance
    for Web Applications
    https://msdn.microsoft.com/en-us/library/bb924356.aspx
    

    View Slide

17. WHAT ABOUT THE CLIENT?
    

    View Slide

18. PERFORMANCE TAB
    

    View Slide

19. NETWORK TAB
    

    View Slide

20. View Slide

21. MEANWHILE ON THE SERVER…
    

    View Slide

22. SYMFONY WEB PROFILER
    

    View Slide

23. View Slide

24. PROFILING AS A SERVICE
    

    View Slide

25. BLACKFIRE
    

    View Slide

26. View Slide

27. View Slide

28. BLACKFIRE PLUGIN
    

    View Slide

29. BLACKFIRE PLUGIN
    

    View Slide

30. $config = new \Blackfire\ClientConfiguration(

    $clientId, $clientToken

    );
    $client = new \Blackfire\Client($config);
    $probe = $client->createProbe();
    // code to be profiled
    $client->endProbe($probe);
    BLACKFIRE SDK
    

    View Slide

31. TYPICAL PROBLEM GROUPS
    I/O WAIT
    CPU USAGE
    MEMORY
    NETWORK LATENCY
    SQL QUERIES
    

    View Slide

32. HOW TO DEVELOP FOR
    PERFORMANCE?
    DON’T!
    

    View Slide

33. DEVELOP FOR
    READABILITY
    MAINTAINABILITY
    SAFETY
    

    View Slide

34. WHAT DOES THAT MEAN?
    SOLID
    YAGNI
    DRY
    

    View Slide

35. AVOID
    ABSTRACTIONS & MAGIC

    You are solving concrete business problems
    MICRO-OPTIMIZATIONS

    Saving a few µs in cpu cycles never matters
    GUESSING

    Only real data from your performance counts
    

    View Slide

36. View Slide

37. PHP 7
    

    View Slide

38. http://www.zend.com/en/resources/php7_infographic
    

    View Slide

39. PHP INSPECTIONS (EA EXTENDED)
    

    View Slide

40. BYTE CODE CACHE
    APC
    OpCache
    WinCache
    

    View Slide

41. FOR SYMFONY

    opcache.max_accelerated_files = 20000
    realpath_cache_size=4096K
    realpath_cache_ttl=600
    https://symfony.com/doc/current/performance.html
    

    View Slide

42. AUTOLOADING
    Use composer classmap
    composer dump-autoload --optimize \
    --classmap-authoritative --no-dev
    composer install -noa
    

    View Slide

43. VARNISH
    https://varnish-cache.org
    

    View Slide

44. Varnish Cache is a web application accelerator
    also known as a caching HTTP reverse proxy.
    You install it in front of any server that speaks
    HTTP and configure it to cache the contents.
    Varnish Cache is really, really fast. It typically
    speeds up delivery with a factor of 300 - 1000x,
    depending on your architecture.
    https://varnish-cache.org/intro/index.html#the-basics
    

    View Slide

45. PHP
    NGINX
    VARNISH
    NGINX
    NGINX
    

    View Slide

46. 
    

    View Slide

47. CACHEABLE HTTP RESPONSES
    Cache-Control
    Expires
    ETag
    Last-Modified
    

    View Slide

48. UTILITIES
    SYMFONY
    Helper methods in Response-object

    Controller-Annotations @Cache
    PSR-7

    https://github.com/micheh/psr7-cache
    

    View Slide

49. CODE SAMPLE
    PSR-7 Response
    $response->withHeader(
    ‘Cache-Control’,
    ‘public, s-maxage=600, must-revalidate’
    );
    

    View Slide

50. https://developers.google.com/web/fundamentals/performance/optimizing-content-efficiency/http-caching
    

    View Slide

51. APPLICATION CACHING
    

    View Slide

52. DON’T BE FOOLED
    REDIS IS NOT EASY TO SETUP
    

    View Slide

53. APCu
    FILESYSTEM
    DATABASE
    

    View Slide

54. ⚠
    USE WITH CPU-HEAVY TASKS
    NOT MEMORY-HEAVY TASKS
    

    View Slide

55. DOCTRINE
    

    View Slide

56. http://docs.doctrine-project.org/projects/doctrine-orm/en/latest/reference/
    improving-performance.html
    METADATA & QUERY CACHES

    it is strongly discouraged to use Doctrine without a Metadata and Query cache
    (preferably with APC or Memcache as the cache driver).
    ALTERNATIVE RESULT FORMATS

    Make effective use of the available alternative query result formats like nested
    array graphs or pure scalar results, especially in scenarios where data is loaded
    for read-only purposes.
    

    View Slide

57. View Slide

58. https://github.com/igorw/retry/issues/3
    

    View Slide

59. BACK IT UP WITH DATA!
    MEASURE
    ANALYZE
    OPTIMIZE
    MEASURE AGAIN
    

    View Slide

60. THANK YOU!
    

    View Slide