Rolf Huisman presented on optimizing DevOps and automation workflows. He discussed moving from traditional DTAP environments to a risk-based OTA-ASP model. Content and COTS artifacts require separate workflows from code. Machine learning models need to consider data privacy regulations. Patching can be automated using APIs. Tests and checks should only be used if they provide sufficient value relative to the time required. Feedback loops can be optimized by reducing context switches and only building code initially. The goal is to optimize risk, time and number of feedbacks to enable pushing to production in under 60 seconds.
2. DEVOPS &
AUTOMATION
Disclaimer
This talk contains some controversial
approaches.
Use with care and at your own risk.
My opinions are my own and not my
employers ☺
12. DEVOPS &
AUTOMATION
Do companies still use these ?
FORTRAN (1957), COBOL (1959),
Python (1990), Java (1995), C# (2000)
While Word documents go to die,
Source code will outlive;
employments, companies, and people
13. DEVOPS &
AUTOMATION
… as Code
Configuration as Code
Infrastructure as Code
Acceptance as Code
Architecture as Code
Policies as Code
15. DEVOPS &
AUTOMATION
Agenda
Code from idea to production
Non-human code
Machine learning and AI
Patching
Non-code artifacts
Content
COTS
Optimize the flow
41. DEVOPS &
AUTOMATION
Merging Infra and Dev
Artifact
Application
Required Infra
Test Automation
(Deploy process*)
All can be partly 3rd party
Base OS
(DSC)
Application + OS
(Exe + DSC)
Monitoring pack
(Dll)
42. DEVOPS &
AUTOMATION
Artifact
Version A:
Of the application
Needs Version B
Of the stack
Needs Version C
To test it
Artifact
Application
Required Infra
Test Automation
(Deploy process*)
All can be partly 3rd party
49. DEVOPS &
AUTOMATION
Machine learning development
Data Scientist
Code
Data
Prediction
StoriesStoriesStoriesStories Master
Data
Algorithm
Code e.g.
R, Python, ONNX
50. DEVOPS &
AUTOMATION
Machine learning development
Data Scientist
Code
Data
Prediction
StoriesStoriesStoriesStories Master
Data
Algorithm
Master Data
Management
Master Dataset
Version Repository
MDM data set ID
Algorithm code
(Machine model)*
Prediction Tests
Code e.g.
R, Python, ONNX
51. DEVOPS &
AUTOMATION
Have a talk about GDPR (AVG)
Are models PII data ?
Remove a single person from a model ?
52. DEVOPS &
AUTOMATION
Have a talk about GDPR (AVG)
Are models PII data ?
Remove a single person from a model ?
No consensus yet.
Be prepared to rewrite history
59. DEVOPS &
AUTOMATION
Patching (Example office)
Configuration BaseOfficeMachines
{
Node 'NodeName'
{
xHotfix HotfixInstall
{
Ensure = "Present"
Path = "http://hotfixv4.microsoft.com/.../358323_intl_x64_zip.exe"
Id = "KB2937982"
}
}
}
Tip:
Make two commits;
1. Absent
2. Present
Allows you to test
and roll back
quicker and easier
70. DEVOPS &
AUTOMATION
Handling content
While you can combine it in the pipeline.
Do you want to ?
• Content approval hindering patches.
• Content changes pushing unintended
functionality
Usually ends up in delays or branch hell.
74. DEVOPS &
AUTOMATION
Example BPM Tool A
Technical
Development
Test & Acceptance
Business
Authoring
Staging
Publishing
Production
Java Script processors
Upgrade testing
Custom reports
Custom validators
Deployment scripts
75. DEVOPS &
AUTOMATION
Example BPM Tool A
Technical
Development
Test & Acceptance
Business
Authoring
Staging
Publishing
Production
Authoring New
Processes
Review in staging
Publishing means
corporate standard
76. DEVOPS &
AUTOMATION
Example BPM Tool A
Technical
Development
Test & Acceptance
Business
Authoring
Staging
Publishing
Production
Authoring New
Processes
Review in staging
Publishing means
corporate standard
Export
Import
79. DEVOPS &
AUTOMATION
COTS
Commercial of the Shelf (COTS)
“My tool/vendor
does not support it”
Artifact
Application
Required Infra
Test Automation
(Deploy process*)
83. DEVOPS &
AUTOMATION
COTS
Service and patch requirements:
“Patches and Releases are enabled and
supported by the supplier to be able to be;
released, deployed, and rolled out without
human intervention using an automated
process using the continuous delivery
pipeline within the agreed timeframe (…)”
84. DEVOPS &
AUTOMATION
COTS
For content applications:
“Supplier commits to be able to support
export, merge, and import of all models,
processes and data using; command line
tooling or rest api’s, to facilitate
automated content delivery processes,
across environments.”
87. DEVOPS &
AUTOMATION
Optimize the flow
Suppose you now have that landscape
But it keeps growing and becomes slow
Thought experiment:
“What would happen if we would just go
to production directly”
89. DEVOPS &
AUTOMATION
What is your risk ?
Reality:
What can you get away with ?
What will your customers tolerate ?
What does your risk budget handle ?
90. DEVOPS &
AUTOMATION
What is your risk ?
Reality:
What can you get away with ?
What will your customers tolerate ?
What does your risk budget handle ?
So… what can we do ?
118. DEVOPS &
AUTOMATION
Takeaways
1. “As code” implies code workflows
2. Code does not always imply humans
3. Don’t force code workflow on business
4. By knowing your risk, you can optimize
119. DEVOPS &
AUTOMATION
So let us speed up DevOps
Rolf Huisman
IT Architect @ InfoSupport
@rlrhuisman
Rolf.Huisman@infosupport.com
https://www.linkedin.com/in/rolfhuisman/
120. DEVOPS &
AUTOMATION Do you want to gain more
knowledge about Microsoft
technology?
The Future Ready Skills program
offers online courseware, online
labs, live Q&A’s and expert
sessions, so you can acquire
your official Microsoft Certificate
in the most efficient way.
For more information:
aka.ms/frsblog
FUTURE READY
SKILLS