Overview

A brief history of the DevOps movement and community with a cross section view of methods and practices that break down the barriers to business agility and continuous delivery of applications.

For a term less than eight years old, DevOps remains a mysterious pursuit for many software engineering organizations.

DevOps seems intangible: you can't buy it, there is no certification, and no universal definition, yet everyone who wants it or does it has trouble identifying it.

We will cover these topics so that you can understand and chart your own journey to DevOps.

• A Definition and Cultural Rendering of DevOps
  • A Brief History of DevOps
  • DevOps Success, Challenges, and the Journey
• The Technical Journey to DevOps
  • Agile Infrastructure and Infrastructure as Code
    • Microservices
    • Pets versus Cattle versus Bacteria
    • Test, Build, Deploy Pattern
  • Infrastructure Orchestration and Models
• Epilogue: BusinessOps
  • We are all DevOps: your call to action!

Currently:

$ cat ~/.signature

DevOps and Automation Architect, Nutanix || mark.lavi@nutanix.com
mobile:+1-650-400-2100 || Twitter @calm_mark || GitHub @mlavi

$ cat ~/.profile || curl http://mlavi.github.io/about/

Previously:

• DevOps Lead at Pertino (Cloud VPN provider)
• Operations at Kaazing (HTML5 WebSocket pioneer)
• Webmaster at Netscape.com, cnnfn.com, Silicon Graphics
  • Technology Evangelist for LDAP and JavaScript

Infrastructure as Code blog = http://mlavi.github.io

A culturally rendered term, but Mark's time tested definition¹ follows:

DevOps is the process of removing all friction
between the developer and customer value.

DevOps has many implications (values, tools, and practices) and it is dynamically bound to the capabilities of the people who practice it, therefore it can vary person by person on the same team.

It is a bi-directional process and concerns all internal and external customers. e.g.: DevOps for yourself on your laptop.

• Illustration: DevOps Automation Diagram
• Blog: I Dream of DevOps, but What is DevOps?

¹Portmanteau: two or more terms combined; a hybrid or mash-up

DevOps is the process of removing all friction
between the developer and customer value.

Ask yourself: Does [this thing] meet our definition of DevOps?

Use your definition as a lens to evaluate any use of the term!

• Blog: Why is DevOps So Hard?

2009:

• February: Agile System Administration Google Group = virtual community begins
• May: Velocity conference in San Jose, CA:
  • "Agile Infrastructure" Slides by Andrew Clay Shafer
  • "10+ Deploys Per Day: Dev & Ops Cooperation at Flickr" Video Slides by John Allspaw and Paul Hammond
• August: Agile 2009 conference in Chicago
  • Patrick Debois rejected talks leads him to find Andrew Clay Shafer.
    Discussion ensues and the term DevOps is born!
• October: DevOpsDays Belgium = physical community begins

² The History Of DevOps by Damon Edwards,
   DevOps: A History by Nell Shamrell-Harrington

• 2013: The Phoenix Project book by Gene Kim, Kevin Behr, George Spafford
  • A parable that illustrates the business impact of DevOps on a traditional manufacturer
• 2014: Annnual State of DevOps Report begins
  • "...high-performing IT organizations are twice as likely to exceed their profitability, market share and productivity goals, and that high performers achieved higher levels of both throughput and stability."

• 2016: Measuring DevOps ROI
  • 200X deploy applications more frequently than slow performers
  • 2,555X faster lead times
  • 24X faster recovery times
  • 3X lower change failure rates
  • 4600: survey participants
• Today: a growing mysterious world-wide movement, buzzword, and dynamic community
  • "Rediscovery" of lean processes, Agile Manifesto, etc.
  • Software discipline applied to refactoring infrastructure and operations
  • Application architectures refactor around operations
  • Manufacturing theory applied to technical work

Success

Technology + Culture transformation for business agility³

Challenges

• Fragmentation: ecosystem involves multiple tools and platforms, fragile integrations
• People: DevOps superstars bridge gaps, but hiring is difficult, expensive, and unscalable

³ Blog: Why is DevOps so Hard?
   Manufacturing Analog: Toyota Production System

We are all DevOps: continual evolution for organizations

• Cultural change to reduce silos between Developer + Test + Operations
  • Dev + Ops; not developers vs. test/QA vs. operations
  • e.g.: ship a new feature by the end of the quarter to satisfy PR and marketing campaigns
• Iteration and automation yielding:
  • Distributed Work: domain expertise democratized, repeatable, and auditable
    • Continuous integration, continuous delivery, continuous deployment
  • Agility: everything is ephemeral; no single point of failure
    • Fail fast, fix fast = antifragile attitude, minimize risk, increase flow

• Closed loop feedback for health and value measurements⁴
  = monitors + logs + metrics for KPIs: Key Performance Indicators
  • Systems yeild domain KPIs which should roll up to business KPMs
  • Continuous Feedback = closed loop operations
    • What is the health of our application, APIs, key performance metrics?
    • Are we increasing the velocity and value in our pipelines?
• All become developers and evolve into service engineers

⁴Proverb: "What is measured improves."

• Agile Infrastructure
• Infrastructure as Code
  • Immutable Infrastructure
  • Microservices
  • Pets versus Cattle versus Bacteria
• Test, Build, Deploy Pattern
• Infrastructure Orchestration and Models

• All services, platforms, and tools are evolving
  • RESTful APIs are everywhere
  • Opportunity to codify and automate everything
  • ...on private, public or hybrid cloud architectures and bare-metal
• Domain expertise democratized, repeatable, and auditable
• Ephemeral everything = agility

Infrastructure as Code = software engineering practices applied to infrastructure

• Version everything: even the database⁵
• Configuration Management⁶
  • Deploy all the things!
  • Build + configure at run-time
• Ephemeral stacks and environments
  • Test all the things!
  • Challenge: full stack orchestration

⁵ Blog: Database Change Management
⁶ Software Configuration Management Systems such as Puppet, Chef, Salt, Ansible, Juju, CFengine, etc.

• Immutable Infrastructure on the rise (cattle)
  • Dynamic versus static configuration, feature flags
  • Bespoke, hand-crafted, snowflake servers = dark magick = friction
  • Automation friction = technical debt = a bug
• Application expertise democratized, repeatable, and auditable
• Operations as Code via behavior and test driven operations
  • I am searching for others to discuss this topic!

• Think of a system that has only:
  • a read-only filesystem
  • environment variables for dynamic, runtime configuration
• Infrastructure artifacts = build system + configuration management at build time
  • Infrastructure is built and placed in a repository for distribution
• Ideal for non-persistent application tiers:
  • Simplify deployment: ship logs, metrics, etc. off the "box"
  • Roll 'em in and out of the load balancer, measure twice!
  • Industry is addressing persistence, but cloud native apps are also a solution
• Closes the risk and gap between development and production

When monolithic, long-lived infrastructure decomposes, so withers the application architecture.

Refactoring and decomposing the application monolithic codebase into modules, each with a public API (REST), enables independent, continuous:

• delivery = build and deploy
• testing and monitoring

per component, team, feature, etc.

Reference: Amazon REST API manifesto

Cloud agility enables ephemeral fleets

• Pet = uptime of years, named, backup maintenence
• Cattle = numbered, can fail and reprovision anytime
• SGI example: 12 years uptime on IRIX! Yesterday's success = today's failure

This is a lens to evaluate infrastructure and ops; evolve to fleet management + app first design

⁷ Pets versus Cattle attributed to Bill Baker, MSFT

Containers

• Agile, lightweight, smaller, faster VMs:
  • ideal approach for continuous delivery of immutable infrastructure artifacts, especially microservices
  • minimal difference between laptop and production, millisecond activation
  • Bacteria⁸ = lifecycle on the order of seconds: build, run, test, destroy
• Early days: Docker as a tool versus Docker as a platform
  • PERL motto = There is more than one way to do it.
• Production challenges remain: orchestration, health, networking, persistence, dynamic configuration -- many issues solved by adopting a container PaaS with limited flexibility

⁸ Bacteria is the term I learned from Tori Wieldt, New Relic Developer Advocate, seems better than "insect" which also is in use.

Between development and production:

• the differences should be minimized⁹
• troubleshooting output and tools should be no different.

Therefore, development environments SHOULD EVOLVE from fully mocked systems to fully integrated application environments, leading to:

⁹ See Immutable Infrastructure

• Create a minimum viable product test and mock your code
• Begin TDD: red, green, refactor

i.e.:

• Test on your laptop
  • Develop tests to satisfy your testable code
• Build on your laptop
• Deploy on your laptop
  • Make code configurable with tools
  • Repeat for integration, metrics, logs, etc.
  • Commit to repository -> continuous integration & delivery

Reference: BTD Pattern Blog Entry

• Orchestration of the entire distributed system/stack/environment
  • Application management lifecycle: all dependencies and operations
• Continuous Deployment + Upgrades:
  • Blue-Green (Red-Black): parallel population, atomic cutover
  • Rolling vs. canary: incremental cutover with testing
  • Test and measure your operations
    • Close the loop to Continuous Delivery
• New disciplines and opportunities:
  • API version management
  • Dynamic feature roll out: Feature flags/lightness, aka "Death to Staging"
  • RuggedDevOps and SecDevOps, NetDevOps

• Local versus Global Redundancy
  • Load balance everything in a local context
  • Distributed, clustered workload schedulers
  • Global load balancing of clusters
• Automate runbooks and change controls = operations as code
  • ChatOps = democratized operations are agile ops

• DevOps escapes the technology domain
• Apply DevOps to business customers, systems, and processes
• Business agility: everything is ripe for automation!
  • Compliance¹⁰ and Security

¹⁰ Compliance at Velocity

DevOps is the process of removing all friction
between the developer and customer value.

$ cat ~/.signature

DevOps and Automation Architect, Nutanix || mark.lavi@nutanix.com
mobile:+1-650-400-2100 || Twitter @calm_mark || GitHub @mlavi

We are All DevOps: Your Next Steps

• Find: your local DevOps community and join us!
  • DevOpsDays.org, MeetUp, and devopsconferences.com
• Evangelize: sildes on my Infrastructure as Code blog = http://mlavi.github.io/post/devops_demystified/