Load Testing Reinvented for DevOps

I take a look at load testing challenges for DevOps and provide some advice around tools and approaches used by flood.io

Presented at Velocity Conference 2018 San Jose CA

Transcript

1. LOAD TESTING Reinvented for DevOps @tim_koopmans Velocity 2018 CA
2. / lōd test / load test verb gerund or present participle: load testing simulate demand on a system and measure it for performance “she was load testing in production, what could go wrong?”
3. /ˌrēinˈvent / re·in·vent verb gerund or present participle: reinventing change something so much that it appears to be entirely new “he launched a ton of browsers and reinvented load testing”
4. /dəˈveləpməntˌäpəˈrāSH(ə)n / dev·ops verb gerund or present participle: devops No definitions found. Search the web “I do devops”
10. 7,856 miles
11. 14 hours
12. 42 milliseconds
13. 59 milliseconds
14. 440 milliseconds
15. ? seconds
17. WHY? https://reddit.com/r/whatcouldgowrong/comments/8o21n6/drifting_on_ice_wcgw/
18. !
19. ! Performance ! Availability ! Reliability ! Scalability
20. Performance
21. Performance
22. Performance
23. Performance
25. Availability
26. Availability
27. Availability
28. Availability
29. Availability
30. Reliability
31. Reliability
32. Reliability
33. Scalability
34. Scalability
35. Scalability
36. Scalability
37. HOW?
38. Simulate demand Measure performance
39. Measure Simulate ?
40. Simulation Options
41. Simulating browsers
42. Simulating protocols
43. flood.io
44. URL testing
45. Caching everywhere
46. Protocol testing
48. So you want to visit amazon.com How many requests will that take? Browser testing
49. 1 Visit +300 Requests Browser Level Protocol Level
50. Browsers are changing
51. Protocols have evolved 1997 HTTP 1.1 2015 HTTP 2.0 2006 XHR jQuery 2011 WebSocket iOS Android 2014 HTML5 ReactJS …
52. Load testing has not
53. Protocol level traps
54. Protocol level recording
55. Protocol level testing
56. Comparing two runs
57. Header differences
58. Cookie differences
59. Session differences
60. POST data differences
61. 6 documents 45 requests 27 static assets 1 XHR 307 header values 39 post values 155 differences 78 cookie values 10 X-CSRF tokens
62. Protocol level debt Lots of code Isolated assets Fragile frameworks
63. Reinventing browser testing
64. Take the best of two breeds + = chrome.flood.io
71. Run it locally
72. Run it remotely
73. Browser level benefits Less code Reusable assets Repeatable frameworks
74. Load Testing Reinvented Simulate User Behavior Don’t pretend to be a browser
75. Use Modern Tools Use browsers not recorders Load Testing Reinvented
76. Leverage Cloud Resources Don’t try this on premise Load Testing Reinvented
77. I do devops
78. I do load testing