DevOps-as-a-Service: Towards Automating the Automation

Accenture Architecture Services
DevOps-as-a-Service:
Towards Automating the Automation
DevOps Enterprise Summit
November 8, 2016
Keith Pleas
DevOps Architecture Sr. Mgr.
keith.pleas@Accenture.com

2
Accenture’s Global DevOps Practice
We support a large variety of industry standard tools – both open
source and vendor/proprietary tools.
250 Clients on Accenture
DevOps Platform
+ 100s more Clients with
DevOps services
4400
DevOps Trained
1700+ DevOps experts
350 DevOps Certified
250+ DevOps
Factory FTE
7 countries
15+Years
heritage of high-
performance CI/CD
& DevOps
2000+ Members of DevOps community
on Yammer sharing strategy, knowledge
& collaboration
759 People
deployed to AT&A
Operations Domain
alone
10%+
Cost reduction
across IT
delivery
New features in hours not days
DevOps Leader
• Accenture developed ADOP DevOps platform (now open-sourced)
• DevOps Managed Services(DCSC)
• Development Lifecycle Management
• Our in-house developed Cloud Platform
• DevOps maturity assessment
• IP/Frameworks: Methods and Tools
Services
• Assessment and Strategy
• Application Architecture
• DevOps Implementation
– Continuous Integration
– Continuous Delivery
– Continuous Testing
• Organizational Change
• IT operating model transformation
• Managed Service
Components
• Build & deployment automation
• Software configuration management
• Test automation
• Environment management & monitoring
• Cloud environment provisioning
• Development & deployment architecture
Key principles
• Test early and often
• Improve continuously
• Automate everything
• Cohesive teams
• Strong source control
• Deliver small increments
• Experiment frequently without regret
Enabling DevOps
Professionals
Global Expert
Network
Courses/Trainings
DevOps Introduction
• Introduction to various DevOps concepts
and how to apply them
• Illustrative examples/case studies
DevOps Academy
• Intensive 2 day classroom course
• Continuous Integration and Delivery
• Infrastructure as Code
Multiple technology specific trainings
Copyright © 2016 Accenture All rights reserved.

DevOps and Accenture
2009 2015201420132012201120102007 200820062005
Agile Conference
Patrick Dubois &
Andrew Clay Shafer
Agile Infrastructure
Qcon: Daniel
Schauenberg
Etsy 50 Deploys
per day
Velocity Conference
John Alspaw & Paul
Hammond: Flikr Deploy
10 time per day
DevOps Days
Belgium #DevOps
Accenture DCSC
Shared Service
Created
2016
Copyright © 2016 Accenture All rights reserved. 3

The Joel Test (August 9, 2000)
A friend from my distant past
• Do you use source control?
• Can you make a build in one step?
• Do you make daily builds?
• Do you have a bug database?
• Do you fix bugs before writing new code?
• Do you have an up-to-date schedule?
• Do you have a spec?
• Do programmers have quiet working conditions?
• Do you use the best tools money can buy?
• Do you have testers?
• Do new candidates write code during their interview?
• Do you do hallway usability testing?
www.joelonsoftware.com/articles/fog0000000043.html

What is DevOps?
Is it culture?

6
Our Challenges: Two Lenses
• Technology
– Automation – Software-defined Infrastructure
– Automate the Automation – Software-defined DevOps
– Templates, tools and stacks
– Enterprise-specific concerns: licensing, security, support
– Technical debt
• People
– Do what you do best - leverage others
– Open Source culture
– New paradigm for engaging Dev and Ops

Focus on Technology
7Copyright © 2016 Accenture All rights reserved.

Roger’s Curve with Moore’s “Chasm”
How innovative are you?

Modern IT methodologies emphasize collaboration
DevOps in context
IDEA PLAN DESIGN BUILD DEPLOY TEST RELEASE OPERATE
DevOps
Continuous Delivery
Continuous Integration
Execution Architecture Development Architecture Testing Strategy Ops Arch
Design Thinking
Lean Manufacturing
Agile Development

10
What about those “non-functional” requirements?
Auditability, consistency, maintainability, modifiability, operability,
recoverability, repeatability, resilience, reusability, robustness,
scalability, security, supportability, scalability, testability, usability
• How are you managing your own continuous delivery artifacts?
• Are all of your build pipeline settings stored in source code control?
• Can you rebuild your pipeline via automation?
• Can you easily scale your pipeline?
• Can you quickly onboard resources and make them productive?
• Could you port your pipeline to a new build environment?
• How are you managing your DevOps technical debt?

Our first “enabler”

12
DevOps Maturity Model – Framework
DELIVERY ORGANISATION AND
METHODOLOGY
• Release Frequency
• Delivery Organisation
• Delivery Process & Methodology
• Metrics & Governance
LEAN DELIVERY GOVERNANCE &
PROCESSES
• Governance Approach
• Continuous Improvement
• Product Management
• Delivery & Change Control
AUTOMATED RELEASE OF SOFTWARE • Source Code Version Control
• Deployment Process
• Deployment Artefact Management
CONTINUOUS INTEGRATION • Traceability
• Continuous Integration
• Code Quality & Unit Test
CONTINUOUS DELIVERY
(INCLUDING AUTOMATED QA)
• Test Strategy
• Test Automation
• Test Data Management
• Deployment Pipelines
AUTOMATED OPERATIONS • Monitoring
• Organisational Considerations
• Resilience
SOFTWARE DEFINED INFRASTRUCTURE
& CLOUD
• Environments
• Infrastructure Suitability
• Responsiveness
PLATFORM/APPLICATION ARCHITECTURE • Architecture
• Impact of Component Failure
• Business Enablement
Architecture
Methodology
Lean Governance & Process
Continuous
Delivery Pipelines
SW Defined
Infrastructure
Automated
Release of SW
Automated
Operations
Continuous
Integration

13
DevOps Maturity Model - Overview
REPEATABLE CONSISTENT OPTIMISEDAD-HOC LEADING2 3 41 5
DELIVERY
ORGANISATION AND
METHODOLOGY
Release scope poorly defined;
subject to catastrophic and ad hoc
change requests
Releases duration exceeds
business need; releases face
disruptive change
Release cadences well defined but
exceeds business need;
Requirements are stable;
Release on demand, multi-speed
releases; time-box meets to
business need (eg: monthly)
Small changes pushed through
the pipeline ; Continuous
deployment enables innovation
LEAN DELIVERY
GOVERNANCE &
PROCESSES
No defined or consistent applied
delivery approach
Delivery approach is well defined,
but largely manually governed
Industrialised project delivery;
governance integrated into
delivery and product
management established
Blame free culture embedded in
governance that uses real-time
data rather than manual inputs
Continuous Delivery / Deployment
supporting agile business change
with automated governance
AUTOMATED RELEASE
OF SOFTWARE
Mostly manual deployments Some CIs automated, environment
tailoring required, no
enterprise tools
Fully automated Single-touch
deployments into environments.
Functioning environments
can be build from nothing
programmatically.
Zero-touch zero-
downtime deployments
CONTINUOUS
INTEGRATION
No consistent use of version
control – builds cannot be traced
back to source code
Source code consistently
managed in VCS; releases
traceable to source
Developers integrate changes by
checking into trunk on regular
basis (daily)
Build is typically green – if build
breaks developers do not make
other changes until resolved
Build is typically green – if build
breaks the CI tooling automatically
reverses the failed change
CONTINUOUS DELIVERY
(INCL AUTOMATED QA)
Fully manual test scripts Testers run a harness / suite Test harness / suite run
automatically for some envs
Automated test suites enforce a
quality gate
Tests run as
functional monitoring
AUTOMATED
OPERATIONS
No monitoring tools Tools in place, but not configured
beyond basic OS checks
Functional monitoring in place and
DR is available
Environment and application
health monitored and
proactively managed
Service level monitoring (perf,
usage) integrated with
infrastructure and
self-healing available
SW DEFINED
INFRASTRUCTURE
& CLOUD
Environments are managed ad-hoc
without consistent blueprint
Environments follow standards but
are manually created
Environments can be created
automatically and follow
common blueprint
SW defined infrastructure that
closely resembles Production
throughout SDLC
Utility based computing that
leverages auto-scaling for several
SDLC aspects
PLATFORM/
APPLICATION
ARCHITECTURE
Big monolithic
application architecture
Architecture based on platforms,
business requires IT for
every change
Resilient Architecture minimising
single point of failures
Modular and scalable architecture,
that allows for some business self
service configuration
Microservice based architecture
supported by end-to-end
ownership in teams

14
DevOps Maturity Model - Overview
REPEATABLE CONSISTENT OPTIMISEDAD-HOC LEADING2 3 41 5

Our second “enabler”

Structure, Standardize, Automate
The complete pipeline to operations with ADOP
IDE
• Eclipse
• X-code
• …
Git / Gerrit
• Source code
repository
• Code branch
management
• Code review
Jenkins
• Continuous
Integration
• Multi-
platform
builds
Code
Analysis
• Code quality
• Coding
standards
rules
• Compliance
Deploy
• Dev app store
• Release
repository
• Environments
Test
• Component
• Behavioural
• Functional
• Security
• Performance
Development
Source
Control
Code
Review
Build
Code
Quality
Deploy Testing
A/B Test in
Production
Please visit github.com/Accenture for more information.
Operations
• Search &
Analytics
• Log Processing
& Analysis
• Visualization

17
Fully automated solution for implementing ADOP
https://hub.docker.com/
Availability Zone
AWS region EU (Ireland)
AWS
AWS cloud
Chef
Server
Public security group
Public subnet
Internet
AWS
AWS Cloud
AWS
CloudFormation
CF template json
VPC
Single ADOP VM
ADOP Provisioning
Gerrit ELK
Nexus
Sonar
Nginx Jenkins
Platform Extensions
Load Cartridge Tomcat
Environment
CloudInit

Fully distributed, highly available, auto-scaling, and self-healing
ADOP platform

ADOP Cartridges
Additional enablers

Dedicated versus Managed ADOP
Dedicated Instance
Single tenancy
Supported by project or by DCSC
Installed on-cloud or on client infrastructure
Supports cartridge/blueprint configuration for
various technologies
Managed Service
Multitenancy
Provided as a managed service by DCSC
Embeds ADLM and Innersource
Costs recovered by service-charge
Hosted primarily in AWS
Configurable to support project requirements
Clients Internal
Projects
Users
Jira/Confluence 148 7 10,800
Innersource 69 2,820 10,600
Usage Jan 2016

Dedicated versus Managed ADOP Tools and Services
Dedicated Instance
On Premise
(Roadmap)
(Beta)
Selenium
AOWP
Digital/01
(Beta)
(Roadmap)
+
Others
AVS
Hybrid
Managed Service
(Roadmap) (Beta)
Compatibility
(Roadmap)

DCSC’s model
• Service Catalog of pre-configured PMO tools
and reporting, enabling rapid mobilization and
low cost run
• A number of processes benefit from a Business
Process Management tool (Activiti), enabling an
automated chasing and reporting
• Services under continuous improvement, rolling
out the latest technology to projects
How to do shared services?
Founding principle for myPMO Services:
Structure....Standardize…. Automate
To reduce local team effort

• The spectrums: standardization, governance, processes
• Consolidating and simplifying the tools portfolio & licensing
• Maturity assessments and coaching clients
• Experiment with ways to obtain better feedback from clients e.g.
periodic surveys, feedback channels, user groups, social media
• Geo-location
• Metrics
• Security, including VPN tunnels across trust boundaries
– Detection (e.g. Security Monkey) rather than imposing restriction
What are the tradeoffs with shared-services?

• Proof of concepts and MVPs
• Onboarding and training
• Starting point for your own DevOps stack
• Starting point for your own shared services
How might you use ADOP?

Focus on People

26
Just one example from Jenkins World…

Benefits of a shared service model
• Re-use of assets between teams
• Sharing of knowledge between teams
• Flexible workloads e.g. person X from project A spends and
afternoon helping person Y on project B when needed
• Creating strong culture that fosters shared values and creates a
great and meaningful place to work
• Being able to have our own culture and identity that we can control
Focus on the People

Lessons learned from a decade of DCSC
• Avoiding the HiPPO (Highest paid person’s opinion)
• Specialization, cross-training, silos, and scalability
• Trust versus Empowerment
– johnclapham.wordpress.com/2016/04/22/inflicting-trust/
• Relationships and staying connected with a lack of face to face
• Ensure internal product teams don’t end up in a silo and lose touch
• Flexibility for mobility of individuals between roles and meet their
needs balanced with staying on engagement through realization
Focus on the People

Where do we go from here?

• Ever-increasing complexity
• Work harder, work smarter, or get help
• Structure, standardize, automate
Evolution: Fact or Theory?

Large media client
Complex DevOps tools landscape
Log Aggregation and Op. Dashboards
Programme and Project Management
Configuration
Management
Test Automation
Requirements
Management
Functional Testing
Continuous Integration
Continuous Delivery
Artefact
Management
Source
Control
Change Management
Defect Management
Continuous Integration/Build Build Orchestration
JenkinsJenkins
Risks and Issues
Continuous
Integration Build
Deployment
Orchestration
Jenkins
Code Validation Functional and API Tests Load Injection
Design DocumentationTeam Dashboards MI and Reporting
Security Tests
Travis CI
AWS
S3
Github
Jenkins
Slack and Lita
Zap
Charles
Box
Composer
BlazemeterPostmanSelenium
Team Communications
Slack
Netsparker
Development and
Prototyping

Let’s look at those NFRs again…
• Versioning, packaging, dependencies
• Chef and Puppet had a head start and showed us the way
• Cloud Formation templates
• Security settings
• Certificates
• ….
Where are our DevOps artifacts?
Auditability, consistency, maintainability, modifiability, operability,
recoverability, repeatability, resilience, reusability, robustness,
scalability, security, supportability, scalability, testability, usability

In general the same way
• Repositories
• DSLs (Domain-Specific-Languages)
– Example: Gradle’s Groovy-based DSL for builds
• Jenkins “Pipeline as Code”- jenkins.io/doc/book/pipeline-as-code/
• A new opportunity for automated testing
How do we manage these?

Ah yes, software factories…did that end well?
“Why I Hate Frameworks” – Benji Smith, September 30, 2005
• A hammer
• A universal hammer
• A hammer factory
• A hammer-factory-building factory
• A general-purpose tool-building factory factory factory
Modification - the fatal flaw of MDA and software factories and…
Automating DevOps
Copyright © 2016 Accenture All rights reserved. 34discuss.joelonsoftware.com/?joel.3.219431.12 34

Putting it all together

• Technology
– Do you have standards for your DevOps practices?
– Do you rationalize your DevOps tools?
– Are you considering your DevOps technical debt?
• People
– Are you attracting or building DevOps talent?
– Are you making the best use of the talent you have?
Does DevOps-as-a-Service make sense to you?
Questions to consider

• www.Accenture.com/DevOps
• ADOP Wiki and quickstart: accenture.github.io/adop-docker-compose/docs/quickstart/
• Mark Rendell DOES15 - www.youtube.com/watch?v=8WRRi6oui34
• “The Joel Test” - www.joelonsoftware.com/articles/fog0000000043.html
• “Why I hate frameworks” - discuss.joelonsoftware.com/?joel.3.219431.12
• DevOps Maturity Model- DevOps@accenture.com
• John Clapham “Inflicting Trust”: johnclapham.wordpress.com/2016/04/22/inflicting-trust/
• Accenture DevOps blog: www.accenture.com/us-en/blogs/blogs-accenture-devops
• Mark Rendell (DCSC) blog: markosrendell.wordpress.com
• Mirco Hering (Maturity Model) blog: notafactoryanymore.com
Resources

DevOps-as-a-Service: Towards Automating the Automation

In this document