Adobe Research

Ideas you will meet

Machine Learning Systems

Recommendations, ranking, feature stores and serving models at scale.

What we already know about this

Our own lessons on these ideas. Read them here before you start, so it makes sense.

• Core Concepts: How Things Actually Work · 9 minDevOps in practice, and where to learn moreWhat DevOps really means, the practices behind it, how to measure delivery performance, and a short reading list.Read hereClose

  It is a culture first

  DevOps removes the wall between people who build software and people who run it. The old model threw code "over the wall" to operations, who feared change; developers wanted change. DevOps aligns both on the same goal: deliver value quickly and reliably. Tools help, but the core is shared ownership, automation and fast feedback.

  The pipeline

  100%

  Figure SDA-CORE-23-1 · The pipeline · © forcsguy.com

  Core practices

  1. Version control for everything: code, configuration and infrastructure definitions.
  2. Continuous integration (CI): merge small changes frequently; each triggers an automated build and tests.
  3. Continuous delivery and deployment (CD): automated, repeatable release to staging and production, with safe rollout (see the deployment strategies lesson).
  4. Infrastructure as code (IaC): define servers, networks and permissions in code (Terraform, CloudFormation, Pulumi, Ansible). Review, version and recreate environments on demand.
  5. Containers and orchestration: package software consistently (Docker) and run it on a platform such as Kubernetes.
  6. Observability: metrics, logs and traces, with alerts on user-visible symptoms.
  7. Security in the pipeline ("DevSecOps"): dependency scanning, secret detection, image scanning, least privilege.
  8. Blameless learning: postmortems and improvement of the system, not of blame.
  9. GitOps: Git is the source of truth for desired state; automation reconciles the real environment with it.

  Measuring delivery performance: the DORA metrics

  Research by the DevOps Research and Assessment (DORA) team found four measures that predict both delivery and business performance:

  Metric	Question
  Deployment frequency	How often do we deploy to production?
  Lead time for changes	How long from commit to running in production?
  Change failure rate	What share of deployments cause a failure needing remediation?
  Time to restore service	How quickly do we recover from a failure?

  The finding that surprised many: high performers are both faster and more stable. Speed and stability reinforce each other when you work in small batches with strong automation.

  Common mistakes

  • Renaming the operations team "DevOps" without changing responsibilities.
  • Buying tools without changing habits.
  • Automating a broken process instead of fixing it.
  • Measuring individuals (lines of code, tickets closed) instead of flow and outcomes.
  • Ignoring security and cost until the end.

  Where to learn more

  These books are widely recommended (buy or borrow them; this course summarises ideas in its own words):

  • Accelerate by Nicole Forsgren, Jez Humble and Gene Kim: the research behind the DORA metrics.
  • The DevOps Handbook by Gene Kim, Jez Humble, Patrick Debois and John Willis: practices and case studies.
  • The Phoenix Project by Gene Kim, Kevin Behr and George Spafford: a novel that explains the ideas through a story.
  • Site Reliability Engineering by Google's SRE team (free online): reliability, monitoring and incidents.
  • Continuous Delivery by Jez Humble and David Farley: pipelines and release engineering.
  • Release It! by Michael Nygard: designing for production failure.
  • Infrastructure as Code by Kief Morris: managing infrastructure with software practices.

  Real-world reading

  Related reading

  • Going faster with continuous deliveryAmazon Builders' Library · Amazon Web ServicesBlogs
  • DevOps Roadmaproadmap.sh community · roadmap.shCourses
  • Screwdriver: Continuous Delivery Build System for Dynamic Infrastructure at YahooYahoo engineers · YahooCase Studies
  • Blue Green DeploymentMartin FowlerBlogs

  Show 6 moreShow fewer

  • AWS DevOps BlogAmazon Web Services engineers · Amazon Web ServicesBlogs
  • Ensuring rollback safety during deploymentsAmazon Builders' Library · Amazon Web ServicesBlogs
  • Software Engineering at GoogleTitus Winters, Tom Manshreck and Hyrum Wright · O'Reilly / GoogleBooks
  • Continuous Integration and Deployment with Bazel at DropboxDropbox engineers · DropboxCase Studies
  • Mozilla Automation Team Engineering BlogMozilla Automation Team engineers · Mozilla Automation TeamBlogs
  • OpenTelemetry conceptsOpenTelemetry · CNCF / OpenTelemetryDocumentation

  Open this lesson with its quiz and exercise