A recap about improving performance in PHP!

Creating performant PHP applications is a long and lonesome road. This talk contains a load of tools and tips to help you make your application Harder Better Faster Stronger!

Transcript

8. DETECTING  THE  PROBLEM
18. TOOLSET
25. THE  PERFORMANCE  LOOP
26. BENCHMARK PROFILE MAKE  CODE   CHANGES
27. OPTIMIZING  PHP
28. Zend  Engine PHPNG
30. Zend  Engine PHPNG Allocates  zval in  HEAP  memory Allocates  zval on  VM  STACK  memory
31. Zend  Engine PHPNG Allocates  zval in  HEAP  memory Allocates  zval on  VM  STACK  memory Pointers to  existing  zvals Creates  new  immutable  zvals
32. Zend  Engine PHPNG Allocates  zval in  HEAP  memory Allocates  zval on  VM  STACK  memory Pointers to  existing  zvals Creates  new  immutable  zvals Lot's  of  reference  counting Avoids  reference  counting
33. Zend  Engine PHPNG Allocates  zval in  HEAP  memory Allocates  zval on  VM  STACK  memory Pointers to  existing  zvals Creates  new  immutable  zvals Lot's  of  reference  counting Avoids  reference  counting Lot's of  garbage  collection Avoids  garbage  collection on  primitive  values
34. Zend  Engine PHPNG Allocates  zval in  HEAP  memory Allocates  zval on  VM  STACK  memory Pointers to  existing  zvals Creates  new  immutable  zvals Lot's  of  reference  counting Avoids  reference  counting Lot's of  garbage  collection Avoids  garbage  collection on  primitive  values No  support   for  JIT  compilers Support   for  JIT  compilers
35. Zend  Engine PHPNG Allocates  zval in  HEAP  memory Allocates  zval on  VM  STACK  memory Pointers to  existing  zvals Creates  new  immutable  zvals Lot's  of  reference  counting Avoids  reference  counting Lot's of  garbage  collection Avoids  garbage  collection on  primitive  values No  support   for  JIT  compilers Support   for  JIT  compilers More  memory  usage Less  memory  usage
36. Zend  Engine PHPNG Allocates  zval in  HEAP  memory Allocates  zval on  VM  STACK  memory Pointers to  existing  zvals Creates  new  immutable  zvals Lot's  of  reference  counting Avoids  reference  counting Lot's of  garbage  collection Avoids  garbage  collection on  primitive  values No  support   for  JIT  compilers Support   for  JIT  compilers More  memory  usage Less  memory  usage SLOWER FASTER
40. THE  COST  OF  IO
41. THE  FILESYSTEM  IS  SLOW
42. $ composer install  -­‐-­‐optimize-­‐autoloader
43. realpath_cache_size realpath_cache_ttl
44. CACHE  YOUR  FRAMEWORK  CLASSES
45. $ composer require  evandotpro/edp-­‐superluminal
46. CACHE  YOUR  SERVICE  CONFIGURATION
47. DON’T  LOG  TO  THE  FILESYSTEM
48. OPTIMIZING  TEMPLATES
49. CACHE  YOUR  TEMPLATES
50. USE  TEMPLATE  MAP
51. EXPLICITLY  SET  VIEW  PATH  IN  CONTROLLER
52. OPTIMIZING  SERVICES
53. AVOID  LOGIC  IN  THE  CONSTRUCTOR
54. LIMIT  THE  NUMBER  OF  GLOBAL  EVENT  LISTENERS
55. USE  LAZY  SERVICES
56. SEND  HEAVY  WORK  TO  THE  BACKGROUND  
57. Caching
59. ENABLE  OPCACHE
60. WRITE  CACHE  TO  CACHE  SERVERS
61. USE  FULL  PAGE  CACHING  WHERE  POSSIBLE
62. CONFIGURE  BROWSER  CACHE  FOR  STATIC  FILES
63. CACHE  FREQUENTLY  USED  DATA
64. CACHE  DATA  FROM  INTEGRATIONS
65. ALSO  CACHE  SERIALIZED  REPRESENTATIONS  OF  AN  OBJECT
66. CACHE  ANNOTATIONS
67. MICRO  IMPROVEMENTS
68. KNOW  YOUR  ROUTER
69. DON’T  USE  DYNAMIC  ASSET  MANAGERS
70. CREATE  STANDARD  FILES
71. OPTIMIZING  DOCTRINE
72. AVOID  PROXIES  IN  OVERVIEWS
73. {%  for  employee  in  employees  %} <tr> <td>{{  employee.name }}</td> <td>{{  employee.company.name }}</td> </tr> {%  endfor %}
74. AVOID  BI-­‐DIRECTIONAL  RELATIONS
75. USE  DTO's  WITH  THE  DATA  YOU  REALLY  NEED
76. SELECT  NEW CustomerDTO(c.name, e.email, a.city) FROM  Customer c   JOIN  c.email e   JOIN  c.address a
77. DON’T  CHANGE  DATA  IN  DOCTRINE  EVENT  LISTENERS
78. AVOID  DQL  AND  THE  ORM  FOR  COMPLEX  QUERIES
79. USE  METADATA  CACHE
80. USE  QUERY  CACHE
81. USE  RESULT  CACHE
82. ENABLE  2-­‐LEVEL  CACHE  FOR  STATIC  DATASETS
83. FLUSH  ENTITIES  IN  BATCH
84. CLEAR()  THE  ENTITY  MANAGER  AFTER  TRANSACTION
85. OPTIMIZING  MySQL
86. AVOID  BLOB  /  TEXT  TYPES
87. OPTIMIZE  COLUMN  SIZE  BY  IMPROVING  DATA  TYPES  /  LENGTH
88. USE  UUID  INSTEAD  OF  INCREMENTAL  IDs
89. ONLY  SELECT  THE  DATA  YOU  NEED
90. ALWAYS  QUERY  THROUGH  INDEXES  AND  JOINS
91. AVOID  HEAVY  SUBQUERIES
92. EXPLAIN  |  DESCRIBE  <query>
94. ENABLE  SLOW  QUERY  LOGGING  DURING  DEVELOPMENT
95. USE  A  BIG  INNODB  BUFFER  POOL  SIZE
96. SKIP  DOMAIN  NAME  RESOLVING  AND  CONNECT  ON  IP
97. KEEP  PRIVILEGES  TABLE  SIMPLE
98. ACTIVATE  QUERY  CACHE