Using Monoids for Business Metrics

Presentation given at TW Geeknight Chennai, Octobeer '17.
Video can be found at https://www.youtube.com/watch?v=RJepu3sbmkU

Transcript

1. Using Monoids for Business Metrics Ashwanth Kumar Principal Engineer @_ashwanthkumar
2. wassup?
3. Business Metrics
4. Everyday Rollups
5. Unique Commits per day Source - github.com/ashwanthkumar
6. we saw aggregates
7. we saw aggregates & unique counts
8. we saw aggregates & unique counts rolled up at various intervals
9. we didn’t see them work at Github scale
10. aggregates (aka) counts
11. counts
12. counts
13. + = counts
14. + = counts
15. counts @ scale
16. uniques
17. uniques
18. uniques
19. U = uniques
20. U = uniques
21. Uniques @ scale Host 1 Host 2 Host 3 Host 4
22. Uniques @ scale Host 1 Host 2 Host 3 Host 4 Reduce at individual hosts
23. Uniques @ scale Master Host * Not the right way to perform uniques, but the simplest
24. Uniques @ scale Master Host
25. we saw how aggregates & unique counts can be computed at scale
26. Wouldn’t it be nice if we can express both kind of aggregations in a consistent way?
27. Monoid is the hero we need & deserve!
28. An operation ( . ) is considered a monoid if: (x . y) . z = x . (y . z) (associativity aka semigroup) identity . x = x . identity = x (identity) trait Semigroup[T] { def plus(left: T, right: T): T } trait Monoid[T] extends Semigroup[T] { def zero: T } monoid
29. Monoids being associative & sometimes commutative are exploited for massive parallel processing
30. ➔ While sum can be achieved in constant memory, distinct counts cannot be. ➔ Approximate structures like HyperLogLog can find unique counts in constant memory (under a known error bound). ➔ 2 or more HLLs can be merged and their result is a monoid. approx. monoids
31. approximate stats now is better than accurate stats tomorrow
32. Aggregations in a distributed environment are computed using scatter and gather primitive.
33. Uniques @ scale Host 1 Host 2 Host 3 Host 4 Scatter
34. Uniques @ scale Host 1 Host 2 Host 3 Host 4 Reduce at individual hosts using HLL Scatter
35. Uniques @ scale Master Host Re-Reduce at Master Host using HLL Gather
36. Uniques @ scale Master Host Gather Re-Reduce at Master Host using HLL
37. We built a system using these @indix called Abel
38. distributed abel architecture 1.1.1.1 1.1.1.2 1.1.1.3 A stats.service.ix 1.1.1.1 1.1.1.2 1.1.1.3 DNS based Load Balancing Count(“a”, 1L) Unique(“a”, 1L) Count(“c”, 1L) Unique(“a”, 1L) Count(“b”, 1L) monoid.plus monoid.plus monoid.plus
39. We are looking forward to open source it soon.
40. Thank you! Image credits jdhancock.com