Why Is My App SLOw? Defining Reliability in Platform Engineering • Jez Humble • GOTO 2023

This presentation was recorded at GOTO Aarhus 2023. #GOTOcon #GOTOaar
gotoaarhus.com
Jez Humble - SRE at Google Cloud & Lecturer at UC Berkeley ‪@JezHumble‬
RESOURCES
continuousdelivery.com
github.com/jezhumble
/ jez-humble
/ jezhumble
sre.google/resources
ABSTRACT
Platform engineering is all fun and games until platform customers start complaining about their apps running slowly. Is it the app code or the platform?
This talk looks at how Google’s Serverless SRE team detects platform-level latency regressions before users, measures the impact of regressions, and tracks performance over time. We’ll discuss the limitations of SLOs in this context and how to take a statistical approach that gives a customer-centric picture of the performance of our platform instead. [...]
TIMECODES
00:00 Intro
02:08 Serverless platform is amazing
05:56 "My app is slow"
08:14 The platform is slow
09:29 Total (end-to-end) latency distribution
10:54 Request delivery latency
12:37 Goal
13:39 Reliability in practice
17:09 Applying to the model
18:39 Stationarity
19:31 2-Sigma Technique
24:30 Mechanics
26:41 Overload score
28:08 Impact analysis
29:03 FAQ
31:56 Backtesting
33:21 Limitations
35:13 Other applications
35:20 Streamlined diagnosis
37:06 Approximate cohort A/B testing
37:33 Conclusions
40:02 Outro
Download slides and read the full abstract here:
gotoaarhus.com/2023/sessions/...
RECOMMENDED BOOKS
Nicole Forsgren, Jez Humble & Gene Kim • Accelerate • amzn.to/442Rep0
Kim, Humble, Debois, Willis & Forsgren • The DevOps Handbook • amzn.to/47oAf3l
Jez Humble & David Farley • Continuous Delivery • amzn.to/452ZRky
Jez Humble, Joanne Molesky & Barry O'Reilly • Lean Enterprise • amzn.to/47pcOXD
/ gotocon
/ goto-
/ gotoconferences
#SLO #SRE #ChaosEngineering #Serverless #PlatformEngineering #2Sigma #GoogleCloud #JezHumble #DevOps #Accelerate
Looking for a unique learning experience?
Attend the next GOTO conference near you! Get your ticket at gotopia.tech
Sign up for updates and specials at gotopia.tech/newsletter
SUBSCRIBE TO OUR CHANNEL - new videos posted almost daily.
ruclips.net/user/GotoConf...

[@allanwind295]

You make it sound like the 2 sigma is deviation is something Google came up with by this is control theory concept is from the 1950s (i.e. en.wikipedia.org/wiki/Western_Electric_rules ). There are value here and enjoyed the presentation.

[@JezHumble]

Thanks for providing a reference to the Western Electric rules and for your kind words about the presentation. I did not mean to imply that Google invented the idea of 2 sigma, that of course is just basic statistics (and I refer to its use in other domains). Rather, the application of this technique to platform reliability (and it's certainly possible other people have applied it here too)