Let's Scale Ruby Applications with Love

Transcript

1. LET’S SCALE RUBY APPLICATIONS WITH ❤
2. WHO AM I ?
4. GitHub: berniechiu Twitter: berniechiu_tw Software Engineer @
5. OVERVIEW • Ruby and API • Ruby and Database • Ruby and Network • Ruby and… just Ruby
6. WHAT MATTERS?
7. AVERAGES?
8. –DHH 「1000 requests that return in 5ms, and 200 requests taking 2,000 ms, still get a respectable 170ms average」
9. Only Very Happy & Very Unhappy Credit from : https://www.youtube.com/watch?v=VPYP7oJJPdo
10. TUNE YOUR DESIRED METRICS
11. SO WHAT MATTERS? • Popular endpoints • Slow response times
13. RUBY WEB + API
14. RESTFUL ENTHUSIASM
18. EVEN MORE DATA?
19. • github.com/zdennis/activerecord-import • github.com/jamis/bulk_insert
20. ALWAYS KNOW WHAT RUNS UNDER RUBY
21. RUBY DATABASE + PAGINATIONS
23. Performance Impact
24. For example - Kaminari
28. KEEPING USER NAVIGATIONS INTACT?
29. Prefetched Count
33. ITERATIONS FROM ACTIVERECORD
34. FIND EACH FIND IN BATCHES • Memory saving • Fetch limited rows on each batch/iteration • Won’t lock the whole table
35. BUT!
36. WORKING WITH UUID?
38. UUID V4 IS SORTABLE?
39. Credit from: https://www.postgresql.org/docs/9.6/uuid-ossp.html
40. Credit From: GitLab CE Custom Implementations
41. SPEED REGAIN
42. RUBY LIBRARIES + NETWORK
43. TYPES OF CACHING
44. SKIPPING TO…
45. LOCALLY FETCHED
49. WAIT…FIXED FOREVER FOR FIXED FOR A PERIOD
51. NO CODE CONNECTION IS FAST CODE
52. Connection Takes Time
53. COPE WITH USE CASES
54. RUBY X THREE
55. BENCHMARKING
56. TOOLS FOR PARTIAL CODE https://github.com/evanphx/benchmark-ips https://github.com/jondot/benchmark-ipsa
57. WHY NOT NATIVE • Only few iterations runs with Benchmark.bm or Benchmark.bmbm • Display time factors instead and need to guess a random iteration counts
61. TEST YOUR THEORY
62. TEST YOUR THEORY Constantized One Time Calculation
63. TEST YOUR THEORY
64. TEST YOUR THEORY
65. SEQUEL GEM https://github.com/jeremyevans/sequel
66. EVERYONE CAN CONTRIBUTE
67. FLAME GRAPHS FROM API
68. TOOLS?
69. RACK MINI PROFILER
70. AN ISSUE?
73. HTML
80. VERIFY!
82. LOAD TEST!
83. BEWARE! • APIs might have different result in different params body • Test with the most common set from online metrics or your own tracking
84. 調效能就跟洗浴缸⼀一樣，拔掉 那個塞⼦子，⽔水很快就流光了了
85. HAPPY HACKING
86. QA