UX Design and Education for Effective Monitoring Tools (Monitorama PDX 2017)

Abstract:
I, like many of us, chose to work in infrastructure because I thought it meant I could avoid talking to other people as much as possible. I was wrong. It turns out that a huge amount of work in monitoring is about empowering the rest of your engineering organization to use the tools you develop correctly, quickly, and effectively. A huge amount of our time is spent explaining how to interpret timeseries data and why averaging percentiles is a bad idea. After feedback from our engineers, we embarked on a journey to redesign our internal monitoring tools and understand where people were struggling with the existing system.

In this talk, I will explain how we approached the problem of making concepts like interpolation, aggregation, and alerting more intuitive and how we identified pain points for new users. I will outline common misconceptions our users have about monitoring and how we cleared up this confusion in our UI without forcing everyone to spend hours on documentation. Rather than copying and pasting existing UX design principles onto our monitoring problems, we will see how we can reinterpret these ideas and apply them to our unique situation to create a better experience for everyone.

Transcript

1. Amy Nguyen @amyngyn Monitorama PDX 2017 UX Design and Education for Effective Monitoring Tools
2. Amy Nguyen @amyngyn Monitorama PDX 2017 Agenda - Background - Motivation - Sharing what you know - Designing what your users want - Recap
3. Amy Nguyen @amyngyn Monitorama PDX 2017 Agenda - Background - Motivation - Sharing what you know - Designing what your users want - Recap
4. Amy Nguyen @amyngyn Monitorama PDX 2017 Hi! - Joined Pinterest in 2015
5. Amy Nguyen @amyngyn Monitorama PDX 2017 Hi! - Joined Pinterest in 2015 - Recent projects:
6. Amy Nguyen @amyngyn Monitorama PDX 2017 Hi! - Joined Pinterest in 2015 - Recent projects: ◦ tracing for performance
7. Amy Nguyen @amyngyn Monitorama PDX 2017 Hi! - Joined Pinterest in 2015 - Recent projects: ◦ tracing for performance ◦ D3 data visualizations
8. Amy Nguyen @amyngyn Monitorama PDX 2017 Hi! - Joined Pinterest in 2015 - Recent projects: ◦ tracing for performance ◦ D3 data visualizations ◦ cache for OpenTSDB data
9. Amy Nguyen @amyngyn Monitorama PDX 2017 Hi! - Joined Pinterest in 2015 - Recent projects: ◦ tracing for performance ◦ D3 data visualizations ◦ cache for OpenTSDB data ◦ so much documentation omg
10. Amy Nguyen @amyngyn Monitorama PDX 2017 Hi! - Joined Pinterest in 2015 - Recent projects: ◦ tracing for performance ◦ D3 data visualizations ◦ cache for OpenTSDB data ◦ so much documentation omg - amynguyen.net @amyngyn pinterest.com/amyngyn
11. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest the world's first visual discovery engine
12. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest - 175 million monthly active users the world's first visual discovery engine
13. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest - 175 million monthly active users - 100 billion pins the world's first visual discovery engine
14. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest - 175 million monthly active users - 100 billion pins - 2 billion boards the world's first visual discovery engine
15. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest - 175 million monthly active users - 100 billion pins - 2 billion boards - 2 billion searches every month the world's first visual discovery engine
16. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest - 175 million monthly active users - 100 billion pins - 2 billion boards - 2 billion searches every month - 150,000 requests served per second the world's first visual discovery engine
17. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest Monitoring the world's first greatest monitoring team
18. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest Monitoring - Graphite, OpenTSDB, Kafka, Storm, Spark, ELK, Sumo Logic, Zipkin the world's first greatest monitoring team
19. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest Monitoring - Graphite, OpenTSDB, Kafka, Storm, Spark, ELK, Sumo Logic, Zipkin - 100 terabytes logged per day the world's first greatest monitoring team
20. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest Monitoring - Graphite, OpenTSDB, Kafka, Storm, Spark, ELK, Sumo Logic, Zipkin - 100 terabytes logged per day - 2.5M metrics ingested per second the world's first greatest monitoring team
21. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest Monitoring - Graphite, OpenTSDB, Kafka, Storm, Spark, ELK, Sumo Logic, Zipkin - 100 terabytes logged per day - 2.5M metrics ingested per second - Over 400 engineers the world's first greatest monitoring team
22. Amy Nguyen @amyngyn Monitorama PDX 2017 About Pinterest Monitoring - Graphite, OpenTSDB, Kafka, Storm, Spark, ELK, Sumo Logic, Zipkin - 100 terabytes logged per day - 2.5M metrics ingested per second - Over 400 engineers customers! the world's first greatest monitoring team
23. Amy Nguyen @amyngyn Monitorama PDX 2017 Our Tools: Dashboards
24. Amy Nguyen @amyngyn Monitorama PDX 2017 Our Tools: Graph Exploration
25. Amy Nguyen @amyngyn Monitorama PDX 2017 Our Tools: Alerting
26. Amy Nguyen @amyngyn Monitorama PDX 2017 Agenda - Background - Motivation - Sharing what you know - Designing what your users want - Recap
27. Amy Nguyen @amyngyn Monitorama PDX 2017 Why should we care about user experience?
28. Amy Nguyen @amyngyn Monitorama PDX 2017
29. Amy Nguyen @amyngyn Monitorama PDX 2017
30. Amy Nguyen @amyngyn Monitorama PDX 2017
31. Amy Nguyen @amyngyn Monitorama PDX 2017 Why should we care about user experience?
32. Amy Nguyen @amyngyn Monitorama PDX 2017 Why should we care about user experience? - Prevent misunderstandings: not everyone is (or should have to be) an expert at interpreting monitoring data
33. Amy Nguyen @amyngyn Monitorama PDX 2017 Why should we care about user experience? - Prevent misunderstandings: not everyone is (or should have to be) an expert at interpreting monitoring data - Developer velocity: help people reach conclusions faster, help your company move faster
34. Amy Nguyen @amyngyn Monitorama PDX 2017 Why should we care about user experience? - Prevent misunderstandings: not everyone is (or should have to be) an expert at interpreting monitoring data - Developer velocity: help people reach conclusions faster, help your company move faster - Data democracy: you don't know what questions people want to answer with their own data
35. Amy Nguyen @amyngyn Monitorama PDX 2017 "The fastest way to become a 10x engineer is to help 10 other engineers do their jobs better. - Wayne Gretzky" - Michael Scott
36. Amy Nguyen @amyngyn Monitorama PDX 2017 Why should we care about user experience? - Prevent misunderstandings: not everyone is (or should have to be) an expert at interpreting timeseries data - Developer velocity: help people reach conclusions faster, help your company move faster - Data democracy: you don't know what questions people want to answer with their own data
37. Amy Nguyen @amyngyn Monitorama PDX 2017 Why should we care about user experience? - Prevent misunderstandings: not everyone is (or should have to be) an expert at interpreting timeseries data - Developer velocity: help people reach conclusions faster, help your company move faster - Data democracy: you don't know what questions people want to answer with their own data
38. Amy Nguyen @amyngyn Monitorama PDX 2017 Why should we care about user experience? Because we can help engineers work correctly, quickly, and independently.
39. Amy Nguyen @amyngyn Monitorama PDX 2017 UX and Your Situation
40. Amy Nguyen @amyngyn Monitorama PDX 2017 UX and Your Situation team
41. Amy Nguyen @amyngyn Monitorama PDX 2017 UX and Your Situation team documentation
42. Amy Nguyen @amyngyn Monitorama PDX 2017 UX and Your Situation team documentation tools
43. Amy Nguyen @amyngyn Monitorama PDX 2017 UX and Your Situation team documentation tools probably paying for vendors TBH?
44. Amy Nguyen @amyngyn Monitorama PDX 2017 UX and Your Situation team documentation tools probably paying for vendors TBH? this talk
45. Amy Nguyen @amyngyn Monitorama PDX 2017 UX and Your Situation team documentation tools probably paying for vendors TBH? this talk things we can all control
46. Amy Nguyen @amyngyn Monitorama PDX 2017 - Background - Motivation - Sharing what you know - Designing what your users want - Recap Agenda
47. Amy Nguyen @amyngyn Monitorama PDX 2017 Sharing what you know
48. Amy Nguyen @amyngyn Monitorama PDX 2017 Sharing what you know 1. Education vs intuition: Don't overload people with too much information.
49. Amy Nguyen @amyngyn Monitorama PDX 2017 will the wifi work? who knows?
50. Amy Nguyen @amyngyn Monitorama PDX 2017 stats.example.metric.errors > 5
51. Amy Nguyen @amyngyn Monitorama PDX 2017
52. Amy Nguyen @amyngyn Monitorama PDX 2017
53. Amy Nguyen @amyngyn Monitorama PDX 2017
54. Amy Nguyen @amyngyn Monitorama PDX 2017 Sharing what you know 1. Education vs intuition: Don't overload people with too much information. 2. Best practices: Use your expertise to determine the most helpful default behavior.
55. Amy Nguyen @amyngyn Monitorama PDX 2017
56. Amy Nguyen @amyngyn Monitorama PDX 2017 Sharing what you know 1. Education vs intuition: Don't overload people with too much information. 2. Best practices: Use your expertise to determine the most helpful default behavior. 3. Potential pitfalls: Make it hard to do the wrong thing.
57. Amy Nguyen @amyngyn Monitorama PDX 2017
58. Amy Nguyen @amyngyn Monitorama PDX 2017 It looks like you're trying to alert on the most recent data. Are you sure you want to do that?
59. Amy Nguyen @amyngyn Monitorama PDX 2017 - Background - Motivation - Sharing what you know - Designing what your users want - Recap Agenda
60. Amy Nguyen @amyngyn Monitorama PDX 2017 Designing what your users want 1. Performance: Do whatever it takes to make it fast.
61. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit!
62. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend
63. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend - Roll-up data over long time ranges
64. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend - Roll-up data over long time ranges - Store latest data in memory (e.g., Facebook's Gorilla paper and Beringei project)
65. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend - Roll-up data over long time ranges - Store latest data in memory (e.g., Facebook's Gorilla paper and Beringei project) - Add a cache layer (e.g., Turn's Splicer project)
66. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend - Roll-up data over long time ranges - Store latest data in memory (e.g., Facebook's Gorilla paper and Beringei project) - Add a cache layer (e.g., Turn's Splicer project) yeah sure get back to us in 6 months
67. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend - Roll-up data over long time ranges - Store latest data in memory (e.g., Facebook's Gorilla paper and Beringei project) - Add a cache layer (e.g., Turn's Splicer project) - Frontend
68. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend - Roll-up data over long time ranges - Store latest data in memory (e.g., Facebook's Gorilla paper and Beringei project) - Add a cache layer (e.g., Turn's Splicer project) - Frontend - Don't reload existing data if the user changes the time window
69. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend - Roll-up data over long time ranges - Store latest data in memory (e.g., Facebook's Gorilla paper and Beringei project) - Add a cache layer (e.g., Turn's Splicer project) - Frontend - Don't reload existing data if the user changes the time window - Prevent the user from requesting the data incessantly
70. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend - Roll-up data over long time ranges - Store latest data in memory (e.g., Facebook's Gorilla paper and Beringei project) - Add a cache layer (e.g., Turn's Splicer project) - Frontend - Don't reload existing data if the user changes the time window - Prevent the user from requesting the data incessantly - Lazy-load graphs on a dashboard
71. Amy Nguyen @amyngyn Monitorama PDX 2017 Performance: Low Hanging Fruit! - Backend - Roll-up data over long time ranges - Store latest data in memory (e.g., Facebook's Gorilla paper and Beringei project) - Add a cache layer (e.g., Turn's Splicer project) - Frontend - Don't reload existing data if the user changes the time window - Prevent the user from requesting the data incessantly - Lazy-load graphs on a dashboard - Disclaimer: We haven't done all of these things.
72. Amy Nguyen @amyngyn Monitorama PDX 2017 Designing what your users want 1. Performance: Do whatever it takes to make it fast. 2. Exploration: Make it easy to try things without fear.
73. Amy Nguyen @amyngyn Monitorama PDX 2017
74. Amy Nguyen @amyngyn Monitorama PDX 2017
75. Amy Nguyen @amyngyn Monitorama PDX 2017
76. Amy Nguyen @amyngyn Monitorama PDX 2017
77. Amy Nguyen @amyngyn Monitorama PDX 2017 Designing what your users want 1. Performance: Do whatever it takes to make it fast. 2. Exploration: Make it easy to try things without fear. 3. Simplicity: Make it easy to figure out what to do.
78. Amy Nguyen @amyngyn Monitorama PDX 2017
79. Amy Nguyen @amyngyn Monitorama PDX 2017
80. Amy Nguyen @amyngyn Monitorama PDX 2017
81. Amy Nguyen @amyngyn Monitorama PDX 2017 manually type in your metric if you know the name somehow
82. Amy Nguyen @amyngyn Monitorama PDX 2017 manually type in your metric if you know the name somehow aggregator
83. Amy Nguyen @amyngyn Monitorama PDX 2017 manually type in your metric if you know the name somehow aggregator hope you remember what tags are available for this metric lol
84. Amy Nguyen @amyngyn Monitorama PDX 2017 manually type in your metric if you know the name somehow aggregator secret bonus: you can downsample?! hope you remember what tags are available for this metric lol
85. Amy Nguyen @amyngyn Monitorama PDX 2017 manually type in your metric if you know the name somehow aggregator secret bonus: you can downsample?! hope you remember what tags are available for this metric lol
86. Amy Nguyen @amyngyn Monitorama PDX 2017
87. Amy Nguyen @amyngyn Monitorama PDX 2017
88. Amy Nguyen @amyngyn Monitorama PDX 2017
89. Amy Nguyen @amyngyn Monitorama PDX 2017 available options you probably don't need to touch
90. Amy Nguyen @amyngyn Monitorama PDX 2017 manual entry for power users
91. Amy Nguyen @amyngyn Monitorama PDX 2017 relevant information needed to create a query!
92. Amy Nguyen @amyngyn Monitorama PDX 2017 Designing what your users want 1. Performance: Do whatever it takes to make it fast. 2. Exploration: Make it easy to try things without fear. 3. Simplicity: Make it easy to figure out what to do.
93. Amy Nguyen @amyngyn Monitorama PDX 2017 - Background - Motivation - Sharing what you know - Designing what your users want - Recap Agenda
94. Amy Nguyen @amyngyn Monitorama PDX 2017 "The fastest way to become a 10x engineer is to help 10 other engineers do their jobs better. - Wayne Gretzky" - Michael Scott
95. Amy Nguyen @amyngyn Monitorama PDX 2017 UX and Your Situation team documentation tools probably paying for vendors TBH? this talk things we can all control
96. Amy Nguyen @amyngyn Monitorama PDX 2017 Sharing what you know 1. Education vs intuition: Don't overload people with too much information. 2. Best practices: Use your expertise to determine the most helpful default behavior. 3. Potential pitfalls: Make it hard to do the wrong thing.
97. Amy Nguyen @amyngyn Monitorama PDX 2017 Designing what your users want 1. Performance: Do whatever it takes to make it fast. 2. Exploration: Make it easy to try things without fear. 3. Simplicity: Make it easy to figure out what to do.
98. Amy Nguyen @amyngyn Monitorama PDX 2017 Thanks!