Operability in Serverless Environments

http://tokyo.serverlessconf.io/

Having your web applications inside serverless environments doesn’t mean forgetting operations. Others have already emphasized what to watch out for in terms of operability in serverless environments. Even though the technology stack, APIs and level of abstractions changed, the first principles in running, operating and maintaining your service for your customers is almost the same.

These concepts applies to your organization supporting your serverless applications as well. For example, without a corresponding organizational transformation, having a serverless team becomes a short-term organizational hack. The interest from technical debit will ask for payment. And suddenly, you realize that you’re back to where you started!

In this talk, I will share some lessons learned from running Platform-as-a-Service Solutions from my experiences as a backend services team and a vendor. There are a lot of similarities since both value propositions to developers and the rest of the organization are similar. Conway’s law is still alive and well when setting up teams to build and run your serverless applications.

Transcript

1. Serverlessͷૢ࡞ੑ Allan Espinosa
2. @AllanEspinosa PaaS Engineer
4. @AllanEspinosa CloudFoundry Experience • ։ൃͱӡ༻ • Private Cloud for other teams • “Platform team” “DevOps team” • ೔ຊͰ̏೥ಇ͖·ͨ͠
5. @AllanEspinosa Engine Yard Cloud Experience • DevOpsαϙʔτΤϯδχΞ • chef-solo 0.6.x, 10.x.x, 12.x.x (new) • PaaS Vendor
6. @AllanEspinosa Container Experience Kubernetes, CoreOS, Deis • DevOpsαϙʔτΤϯδχΞ • buildpacks • All container bugs!
7. @AllanEspinosa Operability without Servers • monitoring • deployment • security • networking
8. –M. Conway, “How Committees Invent,” Datamation, 1968. “… organizations which design systems… are constraint to produce designs which are copies of the communication structures of these organizations.” Conway’s Law @AllanEspinosa
9. @AllanEspinosa Operability on Conway won’t be achieved unless • importance not communicated • no discussion • Ops as second class citizens
10. Platform “Teams”
11. @AllanEspinosa Early Days of • Developer focus • ։ൃ͚ͩʂ • no operation tooling • logs, metrics, instrumentation • fork to achieve ૯࠶ੜ
12. @AllanEspinosa Fork • legal! bureaucracies! • priority mismatch • unaccepted patches ऐ͔ͬͨ͠
13. @AllanEspinosa Organizational changes • OSS first • ࠜճ͠ • ϗϨϯι
14. @AllanEspinosa Community Evolution • Operability (Finally!) • Composability
15. @AllanEspinosa Early Chef (0.6) • early corporate contributor • first Chef Server • mostly developers
16. Early Docker
17. @AllanEspinosa Show HN Resume-Driven Development <Project>, Inc. Raging and Forking Operability Time Visibility
18. “Other” Developers
19. @AllanEspinosa “Other” developers “Just focus on your application” • double-edged sword྆ਕͷ݋ • dev-oriented incentives • communicate operability
20. @AllanEspinosa Support Questions Devs • Our app is down. HELP! • How many instances should my application have? Ops • I debugged XXX, expected was A but got B. Can you help us dig deeper?
21. @AllanEspinosa Container “Adopters” • network is “reliable” • decentralized logging • blaming Docker
22. @AllanEspinosa
24. So Serverless
25. @AllanEspinosa Communication • Provider? • Customers? • Community? • Teammates?
26. –Melvin Conway “… flexibiliy of organization is important to effective design.” Conwayͷ΋ͬͱॏཁͳϙΠϯτ @AllanEspinosa
27. Thank You @AllanEspinosa ςΟʔγϟπ͕͋Γ·͢Αʂ