People want their computer services to operate well, and consequently they often set targets for how reliable they hope to be. In addition to target percentages, percentiles are often used to make it easier to examine the data for most user experiences, while ignoring the outliers. But problems often arise because of a tendency to only think about "nines". All services and user experiences are unique, and it is all too common to ignore important data or stress your systems and humans by trying to be more performant than you need to be. Throw away your nines, and let's find some better numbers to use instead.
Connector Corner: Accelerate revenue generation using UiPath API-centric busi...
Throw Away Your Nines
1. Brought to you by
Throw Away Your Nines
Alex Hidalgo
Principal Reliability Advocate at Nobl9
2. Alex Hidalgo
Principal Reliability Advocate at Nobl9 @ahidalgosre
■ Over a decade of SRE experience
■ Previous work includes food service, retail, game show
host, and DJ
■ Wrote Implementing Service Level Objectives
■ Probably tweets too much
6. Percentiles
■ Primarily a way to talk about frequency distributions
■ Primarily used in tech for handling outliers or tails
■ Especially used for discussing latency
7. Percentages
■ Exist in many different forms
■ Used to think about sets of data
■ Often used in tech to talk about availability or error rates
8. SLOs
■ Service level objectives
■ An approach to measuring reliability
■ A decision making tool
33. Service Level Objectives
■ Measurements used to help us think about our service reliability
■ And we want to be very reliable!
■ So lots of nines!
34. SO MANY NINES
“Two Nines” = 99%
“Three Nines” = 99.9%
“Four Nines” = 99.99%
“Five Nines” = 99.999%
35. What do these nines even mean?
Target Per Day Per Month Per Year
99% 14m 24s 7h 18m 3d 15h
99.9% 1m 26s 43m 97s 8h 45m
99.99% 8.6s 4m 23s 52m 35s
99.999% 0.9s 26s 5m 15s
42. Too Many Nines SLO Targets
■ It’s common to want to be very reliable
■ This locks you into targets that are:
● Unachievable
● Unrealistic
● Don’t match the reality of a service
■ People don’t think about what users actually need
43. Be meaningful in your decisions
Don’t just copy what has come before