Slides for my talk at BCN WordCamp 2016. Improve the performance of WordPress installations by using the right tool at every corresponding level in the technology stack.

1. puerto:22
WORDPRESS
PERFORMANCE TUNING
Javier Arturo Rodríguez
@codehead
javier.rodriguez@puerto22.com

2. puerto:22
ABOUT ME
• Certified System Administrator
• Certified Software Architect
• WordPress enthusiast
• CTO at puerto22.com
@codeheadjavier.rodriguez@puerto22.com

3. puerto:22
PERFORMANCE TUNING

4. puerto:22
TRADEOFFS

5. puerto:22
DIMENSIONS OF
PERFORMANCE

6. puerto:22
Easy
Fast
Inexpensive

7. puerto:22
Easy
Fast
Inexpensive

8. puerto:22
Easy
Fast
Inexpensive

9. puerto:22
Easy
Fast
Inexpensive

10. puerto:22
PICK ANY TWO
Easy
Fast
Inexpensive

11. puerto:22
MEASURE

12. puerto:22
OPTIMIZATION
! "
#
Measure Analyze
Change

13. puerto:22
OPTIMIZATION
! "
#
Measure
Pick a metric

14. puerto:22
OPTIMIZATION
! "
#
Analyze
Advance an
hypothesis
Plan a course
of action

15. puerto:22
OPTIMIZATION
! "
#
Change Act!

16. puerto:22
OPTIMIZATION
! "
#
Measure
Get a new
measurement
of the same
metric

17. puerto:22
OPTIMIZATION
! "
#
Measure Analyze
Change

18. puerto:22
CLIENT VS SERVER
$
!!

19. puerto:22
CLIENT VS SERVER
$
!!
!
! !
!

22. Time To First Byte

25. puerto:22
METRICS!
• Client: Load time
• Developer Tools, Yslow, PageSpeed, Pingdom
• Server: Requests per second
• ApacheBench, siege, httperf

26. puerto:22
APACHEBENCH
This is ApacheBench, Version 2.3 <$Revision: 1706008 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Benchmarking demo2.demoinc.com (be patient).....done
Server Software: Apache
Server Hostname: demo2.demoinc.com
Server Port: 80
Document Path: /
Document Length: 10884 bytes
Concurrency Level: 10
Time taken for tests: 15.467 seconds
Complete requests: 100
Failed requests: 0
Total transferred: 1114376 bytes
HTML transferred: 1088400 bytes
Requests per second: 6.47 [#/sec] (mean)
Time per request: 1546.727 [ms] (mean)
Time per request: 154.673 [ms] (mean, across all concurrent requests)
Transfer rate: 70.36 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 54 67 20.9 60 196
Processing: 555 1430 280.4 1466 1808
Waiting: 439 1314 267.0 1369 1644
Total: 615 1497 277.6 1528 1871
Percentage of the requests served within a certain time (ms)
50% 1528
66% 1621
75% 1680
80% 1736
90% 1796
95% 1843
98% 1871
99% 1871
100% 1871 (longest request)

27. puerto:22
SIEGE
** SIEGE 3.0.8
** Preparing 20 concurrent users for battle.
The server is now under siege.. done.
Transactions: 2000 hits
Availability: 100.00 %
Elapsed time: 59.53 secs
Data transferred: 6.73 MB
Response time: 0.01 secs
Transaction rate: 33.60 trans/sec
Throughput: 0.11 MB/sec
Concurrency: 0.19
Successful transactions: 2000
Failed transactions: 0
Longest transaction: 0.04
Shortest transaction: 0.00

28. puerto:22
HTTPERF
httperf --hog --timeout=5 --client=0/1 --server=demo2.demoinc.com --port=80 --uri=/
--rate=10 --send-buffer=4096 --recv-buffer=16384 --num-conns=100 --num-calls=1
httperf: warning: open file limit > FD_SETSIZE; limiting max. # of open files to
FD_SETSIZE
Maximum connect burst length: 1
Total: connections 100 requests 100 replies 100 test-duration 9.903 s
Connection rate: 10.1 conn/s (99.0 ms/conn, <=1 concurrent connections)
Connection time [ms]: min 2.2 avg 2.8 max 6.8 median 2.5 stddev 0.6
Connection time [ms]: connect 0.7
Connection length [replies/conn]: 1.000
Request rate: 10.1 req/s (99.0 ms/req)
Request size [B]: 77.0
Reply rate [replies/s]: min 10.0 avg 10.0 max 10.0 stddev 0.0 (1 samples)
Reply time [ms]: response 1.8 transfer 0.3
Reply size [B]: header 266.0 content 10884.0 footer 0.0 (total 11150.0)
Reply status: 1xx=0 2xx=100 3xx=0 4xx=0 5xx=0
CPU time [s]: user 3.54 system 6.27 (user 35.7% system 63.3% total 99.1%)
Net I/O: 110.7 KB/s (0.9*10^6 bps)
Errors: total 0 client-timo 0 socket-timo 0 connrefused 0 connreset 0
Errors: fd-unavail 0 addrunavail 0 ftab-full 0 other 0

29. puerto:22
ASIDE: LOG
51.15.54.100 - -
[02/Dec/2016:11:45:27 +0000] "GET / HTTP/1.1" 200 5076 "-"
"Mozilla/5.0 (pc-x86_64-linux-gnu)" "-" "D=11724"

30. puerto:22
SPEED AND EFFICIENCY
• speed: swiftness or rate of performance or action: velocity
• efficiency: productivity without waste
https://www.merriam-webster.com/

31. puerto:22
AREAS OF OPPORTUNITY
• Number of Plugins
• Number/size of Assets
• Database Queries
• API Calls

34. https://premium.wpmudev.org/blog/essential-wordpress-plugins/

35. puerto:22
PLUGINS
• Profile
• Delete unused plugins
• Eliminate redundancy
• When in doubt, delete

36. puerto:22
PROFILE!

37. puerto:22
P³ (PLUGIN PERFORMANCE PROFILER)

38. puerto:22
XCACHE + CACHEGRIND

39. puerto:22
ASSETS
• Concatenate
• Minify
• Compress
• Make cache-friendly
• Offload

40. puerto:22

41. puerto:22

42. puerto:22
DATABASE
• Offload
• Index
• Cache
• Objects
• Queries

43. puerto:22
API CALLS
• Theme optimization
• Opcode Caching

44. puerto:22
APPROXIMATE
STATIC-PAGE SERVING

45. puerto:22
DON’T RE-GENERATE
WHAT YOU CAN CACHE

46. puerto:22
CACHING
• varnish
• memcached/redis
• WP Super Cache
• W3 Total Cache

47. puerto:22
CACHING
• Improved serving time
• Less server load

48. puerto:22
CACHING
WP
Database

49. puerto:22
CACHING
WP
Database
varnish

50. puerto:22
CACHING
WP
Databasememcached
redis

51. puerto:22
W3 Total Cache
CACHING
WP
Database

52. puerto:22
W3 TOTAL CACHE
• Disk caching
• Database caching
• Object caching
• CDN

53. puerto:22
DON’T SERVE ASSETS THAT
OTHERS ARE WILLING TO
SERVE FOR YOU

54. puerto:22
OFFLOADING
• Content Distribution Networks
• Edge Serving
• Edge Caching

55. puerto:22
WP

56. puerto:22
IMG
IMG
IMG
IMG
IMG
WP

57. puerto:22
IMG
IMG
IMG
IMG
IMG
WP

58. puerto:22
WP
WP
WP
WP
WP
WP

59. puerto:22
OFFLOADING
• Overcome single-host bottleneck
• Less server load
• Less bandwidth usage

60. puerto:22
OFFLOADING
HTML
CSS
JS
WOF
IMG

61. puerto:22
OFFLOADING
HTML
CSS
JS
WOF
IMG

62. puerto:22
OFFLOADING
HTML
CSS
JS
WOF
IMG

63. puerto:22
SCALING

64. puerto:22
SCALING STRATEGIES
• Database spin-off
• Scale Vertically
• Scale Horizontally

65. puerto:22
DATABASE SPIN-OFF
WP
Database

66. puerto:22
DATABASE SPIN-OFF
WP Database

67. puerto:22
SCALE VERTICALLY

68. puerto:22
SCALE VERTICALLY

69. puerto:22
SCALE HORIZONTALLY

70. puerto:22
SCALE HORIZONTALLY

71. puerto:22
SCALE HORIZONTALLY
WP WP WPWP

72. puerto:22
SCALE HORIZONTALLY
WP WP WPWP
Database

73. puerto:22
LOAD BALANCING
WP WP WPWP
Load Balancer
Database

74. puerto:22
LOAD BALANCING
WP WP WPWP
Load Balancer
Database

75. puerto:22
ELASTICITY
WP WP
Load Balancer
Database

76. puerto:22
ELASTICITY
WP WPWP
Load Balancer
Database

77. puerto:22
ELASTICITY
WP WP WPWP
Load Balancer
Database

78. puerto:22
FRONT-END CACHING
WP WP WPWP
Cache Cache
Load Balancer
Database

79. puerto:22
DATABASE
• Cache
• Index
• Spin off
• Replication
• Partitioning
Check your slow_queries log!

80. puerto:22
REPLICATION
WP WP WPWP
Load Balancer
Master Replica

81. puerto:22
Partition B
PARTITIONING
WP WP WPWP
Load Balancer
Partition A

82. puerto:22
Partition B
PARTITIONING
WP WP WPWP
Load Balancer
Partition A

83. puerto:22
DATABASE
• HyperDB

84. puerto:22
SERVER STACK
Operating System
Web Server
PHP RE
WordPress
Theme Plugins
Hardware
Database

85. puerto:22
SERVER STACK
Operating System
Web Server
PHP RE
WordPress
Theme Plugins
Hardware
Database
SSD, IOPS
Scaling
Cache
Index
Spinoff
Replication
Partitioning
APC
Xcache
varnish
W3 Total Cache
HyperDB
Networking
Resource allocation

86. puerto:22
DEFAULT STACK
Operating System
Web Server
PHP RE
WordPress
Theme Plugins
Hardware
DatabaseApache
Zend
mod_php

87. puerto:22
ALTERNATE STACK
Operating System
Web Server
PHP RE
WordPress
Theme Plugins
Hardware
Databasenginx
HHVM
fastcgi

88. puerto:22
SEE ALSO
• WordPress Optimization
https://codex.wordpress.org/WordPress_Optimization

89. puerto:22
SUMMARY
• Performance Tuning is a process, not a product.
• Measure. Profile.
• Optimize your WordPress installation prior to scaling.
• Look at the big picture.
• Leverage Caching and Content Delivery Networks.

90. puerto:22
Q&A

91. puerto:22
THANK YOU!
Javier Arturo Rodríguez
@codehead
http://scribd.com/javierrgz/