Keen IO Analytics API Crash Course

Kate Heddleston, software engineer at Runscope on what it takes to build metrics in-house.

Transcript

1. Keen.io API
2. Kate Heddleston Software Engineer Runscope
3. What does it take to build metrics in-house? How we’re using Keen. Why use Keen.io? Keen’s API
4. What does it take to build metrics in-house?
5. Infrastructure UX
6. Infrastructure
7. Skylight Coordinator Kafka Storm ! ! ! ! Spout Bolt Cassandra Java
8. -Tom Dale and Yehuda Katz “One year of research. Six months to rewrite.”
9. Skylight Coordinator Kafka Storm ! ! ! ! Spout Bolt Cassandra Java
10. Keen Testing Monitoring/ validation Data fail-overs and uptime Separate infrastructure
11. New Relic
12. –New Relic Website Marketing Person “New Relic Insights is powered by a highly distributed cloud-hosted event database with an innovative architecture that does not require indexing. The database runs on a super-cluster built to support our thousands of customers — giving you access to a database that can query your big data in seconds.”
13. Data purging: expensive and continuous process. Database tuning. Optimizing for reads vs. writes. Determining which metrics can be pre- aggregated. New Relic Challenges
14. Building metrics infrastructure is really hard.
15. UX
16. Metrics can very easily end up like this:
19. Metrics don’t mean anything unless they are answering clear and specific questions.
20. Why do enterprise customers want metrics?
21. To fill up time in their day because your tool is replacing them. To show their boss the thing they spent money on is useful. Because they have a PhD in statistics and just give me the raw data already.
22. Why do general customers want metrics?
23. To identify outliers, both good and bad. To trouble shoot and identify problems. To answer specific questions they have about their content.
24. Metrics are stories.
25. Data Visualization ! Edward Tufte ! Enrico Bertini ! Cole Nussbaumer ! …
26. Statistics
27. Building metrics UX is really hard.
28. How many hard problems do you want to solve at once?
29. Why use Keen.io?
30. Fast prototyping. They deal with infrastructure while you deal with UX. It’s like paying for another developer (or two) on your team.
31. Keen’s API
32. • Aggregates for a period of time. • Series over a period of time. • Group • Filter • count, sum, percentile, median, average, min, max
33. How we’re using Keen.
35. Focus on UX.
36. Usage • When is our API being used? • How much is our API being used? • Where is our API being used?
37. Performance • Show us the worst queries (95th percentile). • Where is our API slow? • When is our API slow? • How many users experience slowness with our API?
40. Event Collection
42. Timeframe and interval
44. Group by
46. Filters
48. Analyses
51. Building metrics in house is complicated and time consuming. We’re using Keen to bring our customers more insight into what’s going on with their APIs. Keen solves one of your two hard problems for you. Keen’s API is simple and easy to use.
52. Questions
53. Resources • Skylight - Tom Dale and Yehuda Katz • http://www.confreaks.com/videos/3394-railsconf-how-to-build-a-smart-profiler-for-rails • Keen Infrastructure Stack • http://blog.leanstack.io/keen-io-tech-stack/ • Data Visualization Experts • http://www.information-management.com/resource-center/?id=10024000 • New Relic Infrastructure • http://newrelic.com/insights/technology#the-database • http://highscalability.com/blog/2011/7/18/new-relic-architecture-collecting-20-billion-metrics- a-day.html (2011) • Keen API Docs • https://keen.io/docs/data-analysis/