TeamTopologies.com
@TeamTopologies
Frozen DevOps?
The not-so-technical
Last Mile
Manuel Pais
co-author of Team Topologies
11 Apr 2024
@manupaisable
Slide 2
Slide 2 text
Team Topologies
5
Organizing business and
technology teams for fast flow
Matthew Skelton & Manuel Pais
IT Revolution Press, 2019
teamtopologies.com/book
Slide 3
Slide 3 text
Remote Team Interactions
Workbook
7
Using Team Topologies
Patterns for Remote Working
Matthew Skelton & Manuel Pais
IT Revolution Press, 2022
teamtopologies.com/workbook
Slide 4
Slide 4 text
“Team Topologies provides a practical
set of templates for addressing the key
DevOps question that others leave as
an exercise for the student”
- Jeff Sussna
8
Slide 5
Slide 5 text
9
DevOps
Slide 6
Slide 6 text
10
2018 → Early Majority
2019 → Late Majority
2020 → Late Majority
2021 → Late Majority
2022 → Late Majority
InfoQ Trends - “General DevOps”
12
“Success at scale requires optimizing
not for the individual, and not for the
team, but for the wider organization -
the “team of teams”.
Highly evolved organizations have
discovered the patterns that work
well for a fast flow of change”
Why are so many
organizations frozen
in the DevOps middle?
21
Slide 15
Slide 15 text
22
We have Cloud, CI/CD,
SRE & Observability
Mid perf
Slide 16
Slide 16 text
23
“Organizations should not expect to
become highly evolved just because
they use cloud and automation…
They are held back by organizational
structure and dynamics”
Slide 17
Slide 17 text
⏳ ⏳
Blocking
Non-
Blocking
Cloud Infra Deploy
Slide 18
Slide 18 text
25
Slide 19
Slide 19 text
“Devops was always as much about
org structure and incentives as any of
the technical details.”
- Andrew Clay Shafer
26
Slide 20
Slide 20 text
27
Organizations where teams have
strong identities that are well
understood, with clearly defined
responsibilities, are far more likely to
be higher performing
Slide 21
Slide 21 text
29
We have Cloud, CI/CD,
SRE & Observability
Mid perf
Slide 22
Slide 22 text
We have teams with a
clear mission and purpose
that is understood by
other teams
30
High perf
Slide 23
Slide 23 text
31
Everyone collaborates
with everyone else
Mid perf
Slide 24
Slide 24 text
32
“Collaboration is expensive
[for real-time work]…
it’s not scalable or repeatable. ”
Slide 25
Slide 25 text
33
Slide 26
Slide 26 text
34
Slide 27
Slide 27 text
35
Slow feedback loop (weeks/months)
Slide 28
Slide 28 text
36
Slide 29
Slide 29 text
37
“if you simply adopt the team types,
you’re missing one of the key
takeaways... which is paying attention
and being deliberate about teams’
setup and their desired interaction”
Slide 30
Slide 30 text
Team API
38
● Services & software owned by the team
● Wiki and documentation
● Practices and principles
● Communication: tools, channels, patterns
● Roadmap & priorities
Team API
40
● Services & software owned by the team
● Wiki and documentation
● Practices and principles
● Communication: tools, channels, patterns
● Roadmap & priorities
● Current and future team interactions
Slide 33
Slide 33 text
Team A
41
Teams we currently interact with:
Team Name Interaction Mode Purpose Duration
Test Automation
Enabling team
Facilitating Understand test
automation and
data mgmt
examples for iOS
2 months
(from Mar 30
to May 29,
1 day per week)
Monitoring &
Telemetry
Platform team
Collaboration Store and visualize
data on product
features usage
3 weeks
(from Apr 13
to Apr 30,
2h per day)
Slide 34
Slide 34 text
“Collaboration” might be
masking blocking
dependencies
42
🤯
Slide 35
Slide 35 text
Team A
43
Teams we expect to interact with soon:
Team Name Interaction Mode Purpose Duration
Infrastructure
team
Collaboration Change and test
AWS modules to
support our new AI
backend service
1 week ??
(May??? June???)
Slide 36
Slide 36 text
Team A
44
Teams we expect to interact with soon:
Team Name Interaction Mode Purpose Duration
Infrastructure
team
Collaboration Change and test
AWS modules to
support our new AI
backend service
1 week ??
(May??? June???)
Slide 37
Slide 37 text
Team A
45
Teams we expect to interact with soon:
Team Name Interaction Mode Purpose Duration
Infrastructure
team
Collaboration Be able to change
Terraform AWS
modules for
backend services
3 half days
(from May 6
to May 8,
9am-1pm CET)
Slide 38
Slide 38 text
Team A
46
Teams we expect to interact with soon:
Team Name Interaction Mode Purpose Duration
Infrastructure
team
Facilitating Understand
internal Terraform
structure & naming
conventions
1 half day
(on May 6)
Infrastructure
team
Collaboration Be able to change
Terraform AWS
modules for
backend services
3 half days
(from May 13 to
May 15,
9am-1pm CET)
Slide 39
Slide 39 text
Team dependencies tracking
47
https://github.com/TeamTopologies/Team-Dependencies-Tracking
Slide 40
Slide 40 text
48
Everyone collaborates
with everyone else
Mid perf
Slide 41
Slide 41 text
We welcome purposeful
interactions that have
clear goals and duration
49
High perf
Slide 42
Slide 42 text
Case Study
50
“DevOps Over Coffee - Adidas”
Slide 43
Slide 43 text
Case Study
51
Slide 44
Slide 44 text
52
We have fully autonomous
“build & run” teams
Mid perf
Slide 45
Slide 45 text
53
“If cognitive load is left “unbounded”
then… performance metrics will be
negatively affected, preventing teams
from evolving to higher levels”
Slide 46
Slide 46 text
54
Slide 47
Slide 47 text
55
https://www.thoughtworks.com/radar
Apr 2021
Slide 48
Slide 48 text
57
Adopt: Platform
engineering product
teams
“We consider platform
engineering product teams to be
a standard approach and a
significant enabler for
high-performing IT.”
-- ThoughtWorks Tech Radar, Vol.24, p.9
Slide 49
Slide 49 text
60
A valuable platform
reduces the cognitive load
of stream-aligned teams.
Slide 50
Slide 50 text
No content
Slide 51
Slide 51 text
“Cognitive load is the total
amount of mental effort
being used in the working
memory”
- John Sweller
62
Slide 52
Slide 52 text
Intrinsic (skills)
Extraneous (mechanism)
Germane (domain focus)
63
Slide 53
Slide 53 text
Intrinsic
Extraneous
Germane
64
“How do I
deploy this
app to K8s,
again?”
Slide 54
Slide 54 text
(Intrinsic)
] Extraneous [
Germane
65
Slide 55
Slide 55 text
66
extraneous cognitive load
for stream-aligned teams
germane cognitive load
for platform teams
→
Slide 56
Slide 56 text
67
Slide 57
Slide 57 text
68
Slide 58
Slide 58 text
69
“Highly evolved firms use a
combination of stream-aligned
and platform teams as the
most effective way to
manage cognitive load at scale”
Slide 59
Slide 59 text
teamperature.com
70
Slide 60
Slide 60 text
71
Slide 61
Slide 61 text
72
Slide 62
Slide 62 text
73
We have fully autonomous
“build & run” teams
Mid perf
Slide 63
Slide 63 text
We have an ecosystem of
loosely coupled teams that
promotes fast flow &
manageable cognitive load
74
High perf
Slide 64
Slide 64 text
Case Study
75
teamtopologies.com/examples
Slide 65
Slide 65 text
76
Flow of change
Slide 66
Slide 66 text
77
Flow of change
Slide 67
Slide 67 text
78
Flow of change
Slide 68
Slide 68 text
79
Slide 69
Slide 69 text
80
We have a platform team
that solves common
problems for teams
Mid perf
Slide 70
Slide 70 text
81
“the existence of a platform team
does not inherently unlock
higher evolution DevOps”
Slide 71
Slide 71 text
“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
82
Source: https://martinfowler.com/articles/talk-about-platforms.html
Slide 72
Slide 72 text
83
“Successful platform teams treat
their platform as a product. They
strive to create a compelling value
proposition for application teams”
Slide 73
Slide 73 text
But what is
“Platform as a Product”?
84
Slide 74
Slide 74 text
85
A (platform as a) product
is optional to use -
no team is forced to use it
Slide 75
Slide 75 text
“Any time standards, practices,
processes, frameworks, or
architectures become a mandate, I’ve
seen little to no adoption.”
- Courtney Kissler
86
Slide 76
Slide 76 text
87
A (platform as a) product is
carefully designed and
curated for its customers
Slide 77
Slide 77 text
88
A (platform as a) product
evolves to take advantage
of technology changes
Slide 78
Slide 78 text
89
A (platform as a) product
uses modern product
management & follows
the adoption lifecycle
Slide 79
Slide 79 text
90
Happier users (engineers)
No technology bloat
Designed for cognitive load
Slide 80
Slide 80 text
91
“[Highly evolved platform teams]
understand their internal customers
and offer a curated set of
technologies for infrastructure and
development capabilities”
Slide 81
Slide 81 text
92
Slide 82
Slide 82 text
93
Platform engineering is
not replacing DevOps, but
rather enabling its goals
Slide 83
Slide 83 text
94
We have a platform team
that solves common
problems for teams
Mid perf
Slide 84
Slide 84 text
We treat the platform as a
product, a curated
experience for our users
97
High perf
Slide 85
Slide 85 text
Platform as a Product is hard work!
98
● Metrics are stacked against you
● Customer-centricity mindset takes time
● Trust is hard to gain (esp in remote context)
● Adoption lifecycle = “you’re never done”
● Shared services legacy that doesn’t go away
Slide 86
Slide 86 text
Case Study
99
Slide 87
Slide 87 text
100
“the key to escaping the middle
phase is a successful platform team
approach, which makes sense given the
fact that implementing a platform
approach well requires well-defined
team responsibilities and interactions”
○ Platform Value
○ Platform Customers
○ Platform Experience
○ Platform Adoption Cycle
TT25 - Platform as a Product
103
Slide 91
Slide 91 text
Platform Bundle (15% OFF)
Slide 92
Slide 92 text
105
1-day training for platform decision makers
“fantastic course with large amounts of insight”
“an eye opener on the importance of trust in high performing teams”