Achieving Hyper-Productivity through the Use of Microservices and PCF

Achieving Hyper-Productivity through the Use of Microservices and PCF
Achieving Hyper-Productivity
through the Use of Microservices
and PCF
Gregor Zurowski | Thomas Seibert

Who We Are
Software Development @ Mercedes-Benz.io
Thomas Seibert
Lead Architect
Gregor Zurowski
Software Architect

What We Do
Mercedes-Benz on the Web
 End consumer oriented
 Focus on product information
 Rolled out to 50+ markets

Meaning of Productivity

What Does Productivity Mean?
 Output/input
 Lines of Code
 Function Point Analysis
 Working overtime
 Costs as the only kind of measured input
 IT as a cost center vs. the engine of competitive advantage
 Impact on customer behavior/resource binding
 Customer value creation ensures impact
 Where do we allocate resources?
Old World
New World

What Does Hyper-Productivity Mean?
 "Hyper" comes from the ancient greek (ὑπέρ) and means over or above
 Reach a new level of productivity
 Is a 10x development team possible?
 Don't work on the pedal, change gear
 Understanding the processes (systems thinking)
 Value stream mapping
 Elimination of bottleneck
 Measuring impact
 You can't improve what you can't measure
 Metrics mean nothing if they are not actionable

Gaining Productivity by Eliminating Waste
The Seven Wastes of Software Development
 Extra features
 Relearning
 Task Switching
 Partially done work
 Handoffs
 Delays
 Defects
based on Mary and Tom Poppendieck’s work on lean software development

Gaining Productivity by Eliminating Waste
How Can Microservices and PCF Help?
 Extra features
 Relearning
 Task Switching
 Partially done work ✓
 Handoffs ✓
 Delays ✓
 Defects ✓

Approximation Velocity
Speed is Nothing without Control
Medium development velocity,
constantly adjusted direction
Low development velocity,
unadjusted direction
High development velocity,
unadjusted direction

Productivity is Context Dependent
 Several domains (Cynefin framework)
 Obvious
 Complicated
 Complex
 Chaotic
 Disorder/confusion
ComplicatedComplex
Chaotic Obvious
Disorder
based on David Snowden’s Cynefin framework

Productivity in an Enterprise Context
Value Stream Mapping in the Enterprise
Value chain
Functional
maturity
Genesis Custom built Product Commodity
Microservice 1
User focused feature
RDBMSElasticsearch
PCF
IaaS OS
Semantic graph
Upper in the value axis
means: More business
orientation, less
generic functionalityMicroservice 2
Electricity
based on Simon Wardley’s maps

Productivity with PCF and
Microservices

Value Line Proposition of Pivotal Application Service
IaaS
BOSH
Data stores
Service Broker
Service
registry
Applications
value line
managed
by platform
managed by
teams
Message queues
Config
server
Hystrix
dashboard
based on Dekel Tankel’s value line diagram
Router Loggregator
Other PCF
components

Platform Features for Increasing Productivity
Opinionated Platform
Platforms such as PCF are opinionated to reduce complexity by making
assumptions and assuring consistency
 Buildpacks abstract away details and complexities of container construction
 Cloud-native components for service discovery and configuration

Self-Service Provisioning
Self-service provisioning facilities for databases and other middleware services
 Marketplace experience for development teams
 Service brokers provision service instances
 Service bindings provide details for attaching (bind) resources to applications
 Service Broker for AWS further decreases operational burden of managing highly-
available middleware services (storage, messaging, databases)
Service Broker
<<Service Instance>>
Middleware
<<bind>> App<<provision>>
environment
variables

Isolation and Decoupling
Provides isolation and decoupling of teams by
providing
 Process-level isolation with containers
 Permission-level partitioning with orgs and
spaces
<<org>>
<<space>> <<space>>
App App App
Service
<<org>>
<<space>> <<space>>
App
Service ServiceService

Scaling and Monitoring
 Scaling
 Vertical scaling
 Configure memory and CPU via manifest files
 Horizontal scaling
 Configure instance count via manifest or runtime
 Scaling components independently
 Functional decomposition using a Microservice architecture
 Monitoring and metrics
 App Manager provides simplified dashboard view of applications and services
 Hystrix dashboard provide overview of communication within Microservice architecture
 PCF Metrics for insights and troubleshooting

Project Bootstrapping

Introduction
Lessons learned we presented at SpringOne 2017:
"Time needed for onboarding new teams has
initially been underestimated"
How did we address this?
Standardized and automated bootstrapping process for ramping up
new projects

Overview
Product Team
Platform Operations
Project
Initiation
SCM Setup PCF Setup CI/CD Setup
API
Gateway
Setup
100%Generate
code
Commit
& push
first
integration
completed

Operations View | 1
 Process gateway including sign-off by stakeholders
Platform Operations
Project
Initiation
API
Gateway
Setup

Operations View | 2
 Create Git repositories for
 application code
 configuration
Platform Operations
Project
Initiation
API
Gateway
Setup

Operations View | 3
 Create orgs and spaces for all deployment environments
 Create service bindings for shared services
 Configure central config server
Platform Operations
Project
Initiation
API
Gateway
Setup

Operations View | 4
 Setup Jenkins pipelines using Jenkins Seed Plugin
 Create multi-branch build job
 Create one deployment job per environment
Platform Operations
Project
Initiation
API
Gateway
Setup

Operations View | 5
 Add routes for exposing application via API gateway in all environments
Platform Operations
Project
Initiation
API
Gateway
Setup

Product Team View | 1
 Invoke code generator for creating application skeleton
 Generated code contains:
 Minimal set of mandatory dependencies
 PCF manifest files
 Jenkins pipelines
 Java sources with unit tests
 Platform-specific configuration
Product Team
Generate
code
Commit
& push

Product Team View | 2
 A simple commit and push to SCM will…
 automatically build the application
 automatically integrate the application into the dev environment
 Developer can access the application through API gateway or via PCF route
Product Team
100%
first
integration
completed!
Generate
code
Commit
& push

Lessons Learned
 Automate as much as possible (obvious)
 In our case – more specifically:
 Don’t expect developers push changes with the CLI
 Don’t offload implementation of CI/CD pipelines to developers
 Provide a simple and extensible CI/CD library
 "Pipeline as code"
 Aim for self-contained build definitions
 Don’t rely on external tooling
 Application repository contains all dependencies for building the application
 Security and automation requirements might collide

Conclusion

What are the Most Important Factors of Productivity?
 "Individuals and interactions over processes and tools"
 Psychological safety
 You can't engineer culture

Thank You

Achieving Hyper-Productivity through the Use of Microservices and PCF

More Related Content

What's hot (20)

Similar to Achieving Hyper-Productivity through the Use of Microservices and PCF (20)

Recently uploaded (20)

Achieving Hyper-Productivity through the Use of Microservices and PCF