Beyond Ad-hoc Automation To Structured Platforms Bridget Kromhout - - PowerPoint PPT Presentation

▶

Nov 18, 2022 343 likes •763 views

Beyond Ad-hoc Automation To Structured Platforms Bridget Kromhout Pi @bridgetkromhout Bridget Kromhout lives: works: Minneapolis, Pivotal Minnesota podcasts: organizes: Arrested devopsdays DevOps Pi @bridgetkromhout navigating

SLIDE 1 Pi @bridgetkromhout

Beyond Ad-hoc Automation

To Structured Platforms

Bridget Kromhout

SLIDE 2 Pi @bridgetkromhout

Bridget Kromhout

lives: Minneapolis, Minnesota works: Pivotal podcasts: Arrested DevOps

rganizes:

devopsdays

SLIDE 3 Pi @bridgetkromhout

navigating infrastructure choices

SLIDE 4 Pi @bridgetkromhout

Previously, on Platform Tales… Docker in Production: Reality, Not Hype

SLIDE 5 Pi @bridgetkromhout

Deconstructing a monolithic Python/Django app into… (as was the style at the time) Go microservices

SLIDE 6 Pi @bridgetkromhout

Peak load: tens of thousands of requests per second Traffic variance: swings 10-20x throughout the week

SLIDE 7 Pi @bridgetkromhout

SLIDE 8 Pi @bridgetkromhout

Containerized builds

runs tests creates versioned deployable artifacts standardizes development environment

SLIDE 9 Pi @bridgetkromhout

SLIDE 10 Pi @bridgetkromhout

also true for infrastructure as code.

SLIDE 11 Pi @bridgetkromhout

SLIDE 12 Pi @bridgetkromhout

SLIDE 13 Pi @bridgetkromhout

“Great job on those

zero-downtime

blue-green

deployments!”

—no CEO ever

SLIDE 14 Pi @bridgetkromhout

canary deployments

SLIDE 15 Pi @bridgetkromhout

Just because you can… …doesn’t mean you should.

SLIDE 16 Pi @bridgetkromhout

SLIDE 17 Pi @bridgetkromhout

Image credit: Simon Wardley

SLIDE 18 Pi @bridgetkromhout

there has to be a better way

SLIDE 19 Pi @bridgetkromhout

let’s talk platforms…

SLIDE 20 Pi @bridgetkromhout

if you can deploy and operate code in any capacity, you have a platform

pinions optional

SLIDE 21 Pi @bridgetkromhout

SLIDE 22 Pi @bridgetkromhout

SLIDE 23 Pi @bridgetkromhout

SLIDE 24 Pi @bridgetkromhout

Large-scale cluster management at Google with Borg - Verma et al. 2015

“Almost every task run under Borg contains a built-in HTTP server that publishes information about the health of the task and thousands of performance metrics”

SLIDE 25 Pi @bridgetkromhout

why > what

Principles > Practices >Tools

SLIDE 26 Pi @bridgetkromhout

SLIDE 27 Pi @bridgetkromhout

SLIDE 28 Pi @bridgetkromhout

reasonable constraints

SLIDE 29 Pi @bridgetkromhout

Constraints are the contract that allows a platform to keep promises.

SLIDE 30 Pi @bridgetkromhout

Everyone has a platform. What promises can your platform keep?

SLIDE 31 Pi @bridgetkromhout

role based access to resources
run code on demand
coordinate cross-service configurations
route public requests
read and write persistent data
add and remove resources
record internal and external events
isolate resources and failures
measure performance/health
detect & determine failure
recover from failures
day two operations

does your platform promise this?

SLIDE 32 Pi @bridgetkromhout

concourse.ci

your pipeline to continuously deliver microservices

(and win buzzword bingo)

SLIDE 33 Pi @bridgetkromhout

promises

SLIDE 34 Pi @bridgetkromhout

structured contracts determine the promises a platform can keep

SLIDE 35 Pi @bridgetkromhout

simple patterns automated by tooling

SLIDE 36 Pi @bridgetkromhout

“Any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the

rganization's communication structure.”

Mel Conway

SLIDE 37 Pi @bridgetkromhout

12-factor apps need 12-factor ops

SLIDE 38 Pi @bridgetkromhout

Use tools and simple patterns to make doing the right thing the easy thing

SLIDE 39 Pi @bridgetkromhout

“To always be

shipping,

you need a

shipyard”

Bret Mogilefsky of 18F,

n cloud.gov

SLIDE 40 Pi @bridgetkromhout

Thank you!