Graphs: Lessons Learned (with Science!)

A 5-10 minute lightning talk given at ATLRUG on May 9, 2012 in which I evangelize some of the lessons I've learned about the hows, whys, and whats of graphing.

Transcript

1. GRAPHS
2. WILL FARRINGTON S @WFARR W OPS
3. LESSONS LEARNED
4. LESSONS LEARNED WITH SCIENCE!
5. GRAPHS YOU NEED TO HAVE
6. $/MIN
7. HOW MUCH MONEY DOES DOWNTIME REALLY COST?
8. HIGH AVAILABILITY IS GREAT... IF YOU NEED IT.
9. USE DATA. INFORM YOUR DECISIONS.
10. FORECASTED GROWTH
11. YOUR THINGS, UTILIZATION, AND ANYTHING THAT CHANGES
12. IT’S NOT A PROBLEM UNTIL IT’S A PROBLEM
13. BUT
14. THEN IT’S A PROBLEM
15. SO PLAN TO HAVE PROBLEMS
16. PLAN TO DEAL WITH PROBLEMS
17. WHEN THEY’RE ACTUALLY PROBLEMS
18. HOW YOU SHOULD MEASURE
19. MEAN
20. STDEV
21. WHAT IS YOUR DISTRIBUTION OF DATA RELATIVE TO THE MEAN?
22. VARIANCE
23. HOW DIFFERENT ARE YOUR DATA POINTS FROM THE MEAN?
24. LOTS OF VARIANCE IMPLIES UNPREDICTABILITY
25. IT'S OK IF A QUERY IS SLOW AS LONG AS IT IS ALWAYS SLOW
26. IT'S OK IF A QUERY IS SLOW AS LONG AS IT IS ALWAYS SLOW FACEBOOK ENGINEERING
27. PERCENTILE
28. 95 98 99 99.9
29. WHAT ARE YOUR GREATEST OUTLIERS?
30. QUERIES/SECOND MEANS NOTHING IF YOUR EDGE CASES SUCK
31. QUERIES/SECOND MEANS NOTHING IF YOUR EDGE CASES SUCK FACEBOOK ENGINEERING
32. WHY YOU SHOULD MEASURE
33. DATA SHOULD INFORM YOUR DECISIONS
34. FEELING DRIVEN DEVELOPMENT
35. THIS THING “FEELS” FASTER
36. THIS THING “FEELS” FASTER NO
37. DATA DRIVEN DEVELOPMENT
38. MONITOR MEASURE MANAGE
39. DATA DRIVEN DEVELOPMENT YES
40. IT’S LIKE THE SCIENTIFIC METHOD ON REPEAT
43. lulz lullingly provided by @techpickles
44. TOOLS
45. NEWRELIC
46. SCOUT
47. METRIKS https://github.com/eric/metriks
48. LIBRATO METRICS
49. (STATSD) + GRAPHITE
50. WILL FARRINGTON S @WFARR W OPS speakerdeck.com/u/wfarr
51. THANKS