Kubernetes helps us tame sprawling microservices architectures and address increased operational complexity. Kubernetes gives developers abstractions and APIs to deploy and run their services. But there is a price to pay in terms of both the in-house operational expertise required and the learning curve for application teams. The elephant in the room is that to run, maintain and evolve Kubernetes, we likely need a dedicated Kubernetes team. Is the tradeoff between better operational tools and introducing a new dependency layer on the path to production for application teams worthwhile? Are we making life easier for application teams or instead reducing their end-to-end ownership? Regardless of all the technical benefits that Kubernetes undoubtedly brings, team interactions are still key for successfully delivering and running services. We will look at a couple of organizations that have succeeded by focusing on reducing the cognitive load for application teams. Unfortunately, many organizations see Kubernetes as “the” platform, rather than just a technical foundation for a true internal platform. In the worst case, they mandate all teams to adopt Kubernetes, regardless of both the application teams’ and the platform’s maturity levels. Successful Kubernetes adoption requires thinking about what a platform really means and learning which team structures and interactions work well. And evolve them over time.

1. TeamTopologies.com
@TeamTopologies
K8s is Not Your Platform,
It’s Just the Foundation
Manuel Pais
co-author of Team Topologies
UK Cloud Infrastructure User Group, Dec 2020
@manupaisable

2. Team Topologies
2
Organizing business and
technology teams for fast flow
Matthew Skelton & Manuel Pais
IT Revolution Press (Sep 2019)
teamtopologies.com/book

3. 3
Is Kubernetes a Platform?
Team Cognitive Load
Team Interactions
Getting Started

4. Is Kubernetes a Platform?
4

5. 5
Source: https://www.infoq.com/news/2019/03/airbnb-kubernetes-workflow
Melanie’s talk: https://www.infoq.com/presentations/airbnb-kubernetes-services

6. 6

7. Kubernetes “platform”
microservices
ops complexity
7

8. Kubernetes “platform”
deploy & run
abstractions
8

9. 9

10. 10
Still need to...
… sizing hosts
… create/destroy clusters
… update to new K8s versions
… decide on namespaces vs clusters
<insert your fav chore here>

11. 11

12. Who is the provider?
12

13. Who is the provider?
Who is the consumer?
13

14. 14

15. “A digital platform is a foundation of
self-service APIs, tools, services,
knowledge and support which are
arranged as a compelling internal
product.”
– Evan Bottcher, 2018
15

16. “A digital platform is a foundation of
self-service APIs, tools, services,
knowledge and support which are
arranged as a compelling internal
product.”
– Evan Bottcher, 2018
16

17. Kubernetes is not your
platform. It’s the foundation.
17

18. “Create a path of
least resistance.
Make the right thing
the easiest thing to do.”
– Evan Bottcher, 2018
18

19. The hard thing about
platforms is to constantly
evolve & adapt to
new consumer needs.
19

20. Team Cognitive Load
20

21. 21

22. 22

23. 23

24. “Cognitive load is the total
amount of mental effort
being used in the working
memory”
- John Sweller
24

25. Intrinsic
Extraneous
Germane
25
“How are
classes
defined in
Java?”

26. Intrinsic
Extraneous
Germane
26
“How do I
deploy this
app,
again?”

27. Intrinsic
Extraneous
Germane
27
“How do
bank
transfers
work?”

28. Intrinsic (skills)
Extraneous (mechanics)
Germane (domain focus)
28

29. (Intrinsic)
] Extraneous [
Germane
29

30. More: ‘Hacking Your Head’
30
Jo Pearce
(@jdpearce)
https://www.slideshare.net/JoPearce5/hacking-your-head-managing-information-overload-extended

31. Be mindful of your
platform choices’ impact
on teams’ cognitive load
31

32. CaseStudy
32

33. “The best part of my day is when I
update 10 different YAML files to
deploy a one-line code change.”
33

34. “The best part of my day is when I
update 10 different YAML files to
deploy a one-line code change.”
– No One, Ever
34

35. 35

36. Clarify (platform) service
boundaries and provide
abstractions to reduce the
cognitive load on teams.
36

37. CaseStudy
37

38. 38
Source: https://medium.com/@pingles/convergence-to-kubernetes-137ffa7ea2bc

39. 39
Low-level AWS service calls (EC2, IAM, STS, Autoscaling, etc.) from January 2015 to January 2017

40. “We didn’t change our organization
because we wanted to use
Kubernetes, we used Kubernetes
because we wanted to change
our organization.”
- Paul Ingles
40

41. “Kubernetes helps us in a few ways:
- Application-focused abstractions
- Operate and configure clusters to
minimise coordination ”
- Paul Ingles 41

42. 42
Low-level AWS service calls since Kubernetes adoption in January 2017

43. 43
enable stream-aligned
teams to deliver work
autonomously with
self-service capabilities ...
Platform Purpose

44. 44
… in order to reduce
extraneous cognitive load
on stream-aligned teams
Platform Purpose

45. “We wanted to scale our teams but
maintain the principles of what
helped us move fast: autonomy,
work with minimal coordination,
self-service infrastructure.”
- Paul Ingles
45

46. Treat the platform
as a product
46

47. Reliable
Fit for Purpose
Focused on DevEx
47

48. 48
on-call support
service status pages
suitable comms channels
response time for incidents
downtime planned & announced
Reliable Platform

49. 49
prototyping
fast, regular feedback
agile, iterative practices
few(er) services, high(er) quality
skilled product management
Fit for Purpose Platform

50. 50
speak the same language
right level of abstractions for
your engineering teams today
#DevEx Focused Platform

51. 51
2018
Infra platform
started with few
services
First customer
(centralized
logging, metrics,
auto scaling)

52. 52
2018
Infra platform
started with few
services
First customer
(centralized
logging, metrics,
auto scaling)
2019
Started using SLAs
and SLOs, clarifying
reliability/latency/etc
Growing traffic in
platform vs AWS

53. 53
...
Addressed critical
cross-functional
needs (GDPR,
security, alerts +
SLOs as a service)
Adoption by HMMT
(Highest Money
Making Team)
2018
Infra platform
started with few
services
First customer
(centralized
logging, metrics,
auto scaling)
2019
Started using SLAs
and SLOs, clarifying
reliability/latency/etc
Growing traffic in
platform vs AWS

54. 54
...
Addressed critical
cross-functional
needs (GDPR,
security, alerts +
SLOs as a service)
Adoption by HMMT
(Highest Money
Making Team)
2018
Infra platform
started with few
services
First customer
(centralized
logging, metrics,
auto scaling)
2019
Started using SLAs
and SLOs, clarifying
reliability/latency/etc
Growing traffic in
platform vs AWS

55. 55
user satisfaction metrics
(Accelerate metrics for business services, NPS, etc)
Platform Metrics

56. 56

57. 57
user satisfaction metrics
(Accelerate metrics for business services, NPS, etc)
adoption & engagement metrics
(% teams onboard, per platform and per service)
Platform Metrics

58. 58

59. 59
user satisfaction metrics
(Accelerate metrics for business services, NPS, etc)
adoption & engagement metrics
(% teams onboard, per platform and per service)
reliability metrics
(SLOs, latency, MTTR, etc)
delivery metrics
(Accelerate metrics i.e. lead time, deploy freq, etc)
Platform Metrics

60. 60
The success of platform
teams is the success of
stream-aligned teams

61. Team Interactions
61

62. 62

63. 63

64. 64service boundary

65. 65service boundarycognitive load

66. 66
strong collaboration with
stream-aligned teams for
any new service or evolution
Platform Behaviors

67. 67

68. 68
provide support and great
documentation for stable
services
Platform Behaviors

69. 69

70. 70

71. 71

72. 72

73. 73

74. 74

75. 75

76. 76
Source: https://landscape.cncf.io

77. 77

78. Getting Started
with team-centric
Kubernetes adoption
78

79. How well can the team understand the
platform/Kubernetes abstractions they
need to use on a regular basis?
1 - Assess cognitive load
79
github.com/TeamTopologies/Team-Cognitive-Load-Assessment

80. What’s the gap between your
Kubernetes implementation and an
internal digital platform?
2 - Define your platform
80

81. Who is responsible for what? Who is
impacted? How do you collaborate on
new platform internal services?
Collaboration vs X-as-a-Service
3 - Team Interactions
81

82. Zalando Kubernetes at Zalando
Mercedes DevOps Adoption at Mercedes-Benz.io
Twilio Platforms at Twilio: Unlocking Developer Effectiveness
Adidas Where Cloud Native Meets the Sporting Goods Industry
ITV ITV's Common Platform v2 Better, Faster, Cheaper, Happier
MAN Truck & Bus How to Manage Cloud Infrastructure at MAN Truck & Bus
Farfetch UX I DevOps - The Trojan Horse for Implementing a DevOps Culture
More platform examples
82

83. 83
https://techbeacon.com/enterprise-it/why-teams-fail-kubernetes-what-do-about-it

84. Resources
84
teamtopologies.com/remote-first
(links, slides, video)
github.com/teamtopologies
(templates, assessments, etc)

85. Principles & Practices for
Modern Platforms (TT08)
teamtopologies.com/training
4-hour training

86. Thank you!
info@teamtopologies.com
86
Matthew Skelton, Conflux
@matthewpskelton
Manuel Pais, Independent
@manupaisable
Copyright © Conflux Digital Ltd 2018-2020. All rights reserved.
Registered in England and Wales, number 10890964
Icons made by Freepik from www.flaticon.com - used under license