Building BMW Group's Customer Engagement Platform on AWS (AMT305) - AWS re:Invent 2018

© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Building BMW Group’s Customer
Engagement Platform on AWS
A M T 3 0 5
Constantin Gonzalez
Principal Solutions Architect
AWS
Julian Roedig
Principal Solution Architect
BMW Group
Patrick Lanners
Solution Architect
BMW Group

What you’ll get out of this session
An introduction to microservices
Real-world customer experience from BMW Group
Methodology
Architecture overview
Architecture deep dives
Guidelines and resources for your own microservices projects

Rigid
Hard to scale
Hard to deploy
Locked-in
Hard to reuse
Easy to break
Monoliths

Monolith development lifecycle
Developers
ReleaseTestBuild
Delivery pipelineApp

Autonomous
Specialized
Agile
Flexible to scale
Easy to deploy
Freedom
Easy to reuse
Resilient
Microservices

Microservice development lifecycle
Developers Delivery pipelinesServices
ReleaseTestBuild
ReleaseTestBuild
ReleaseTestBuild
ReleaseTestBuild
ReleaseTestBuild
ReleaseTestBuild

But how?
Migration?
Networking?
Service discovery?
Deployment?
Which services?

Introduction
Architecture overview
Architecture deep dives
1
2
3
Conclusion4

Lanners & Roedig, BMW Group | AWS re:Invent 2018
THE PLATFORM CONSOLIDATES THE VEHICLE PRODUCTS AS SERVICES.
~7
Engines
~12
Colors
~14
Wheels
~11
Upholstery
~7
Trims
~90
Optional
Equipment
Many possibilitiesto configureyour dream car
=
OUR PLATFORM CONSOLIDATES THE VEHICLE PRODUCTS AS SERVICES

EXAMPLES OF FRONT ENDS THAT USE OUR CONFIGURATOR PLATFORM
Dealer iPad Configurator bmw.de mini.at bmw-motorrad.de 26+ Systems

UNIFIED CONFIGURATOR PLATFORM: CENTRAL
ENGAGEMENT PLATFORM AS BUSINESS ENABLER
Unified Configurator Platform
Accessories
Product Data
Official
Technical Data
Net-Gross-Price
Calculation
Virtual Garage
Constructability
Check
Data Import Data Generation
Product Data Technical Data …
Consumes API
...
Technical
Calculations
Key facts:
- RESTful API
- Used by ~30 products
- Hundreds of Mio API
calls per month
- Supports all BMW
Group brands and
markets

A MONOLITHIC SYSTEM ISN’T BYDEFINITION A BAD THING
Monolith
Team
Clients
Benefits
- Fast development of a first production-ready MVP
- No need to handle the challenges of a distributed system
Challenges and driving factors for going the next step
- Increased number of requirements led to an increase of
the overall complexity

MICROSERVICES WITH API-FIRST STRATEGY: ENABLES FASTER DEVELOPMENT AND
INCREASEDTRANSPARENCY
Benefits
- Increased transparency for API consumer due to API-
first approach
- Multiple teams can work independently
Challenges and driving factors for going the next
step
- Latency performance to provide our service on a
worldwide basis
- Broader solution scope needed to satisfy business
needs
Teams
Clients
MS
MS
MS
MS
API-First

GAME CHANGER: BRINGINGTHE MICROSERVICES IN A “LIFT-THINK-SHIFT”
APPROACHTO AWS
AWS Elastic Beanstalk
Amazon API Gateway
Amazon ECS + AWS Batch
Amazon RDS
AWS-Cloud
Glassfish 4
Microservice
Software Business Logic
API Specification Transformation into API-GW
Transformation from
ORACLEto Postgres
API-Sec. Auth. Information
Dockerizethe business logic
Specifications
WEB-EAM Master Solution
On-Premises Off-Premises
API Interface (generated)
Authentication Information
Docker + GF4
Microservice DB
Software Business Logic
Rewrite import procedures
from PL/SQLto Java
Microservice DB
Import Job
Docker
Software Import Job

WE DEFINED A SINGLE ENTRY POINT TO OUR SERVICES
AWS
API Gateway
AWS
CloudFront
Amazon
Route 53
AWS
Lambda
AWS
DynamoDB
Elastic Beanstalk Environment
Network Load
Balancer
ELB Subnet A/B App Subnet A/B
Amazon
Route 53
VPC Link
EB Environment
Microservice A
Amazon
EC2
Proxy

OUR BASE TECH STACK FOR MICROSERVICES ON BEANSTALK
Application
Load
Balancer
Amazon
EC2
Auto Scaling Group
Amazon RDS
(Multi AZ)
DB Subnet A/B
Security Group
Security Group
Microservice A
nstalk Environment
d
B App Subnet A/B
Amazon
Route 53
Amazon
EC2
Proxy

Architecture deep dives3
Service discovery
AWS Lambda CI/CD
Offline data processing
Advanced Lambda use case
Environments on demand

AWS
API Gateway
AWS
CloudFront
AWS
Route 53
AWS
Lambda
AWS
DynamoDB
ENHANCED FLEXIBILITYTHROUGH DNS-BASED SERVICE DISCOVERY
Network Load
Balancer
Amazon
Route 53
VPC Link
EB Environment
Microservice A
Amazon
EC2
Proxy

AWS
API Gateway
AWS
CloudFront
AWS
Route 53
AWS
Lambda
AWS
DynamoDB
ENHANCED FLEXIBILITYTHROUGH DNS BASED SERVICE DISCOVERY.
Network Load
Balancer
Amazon
Route 53
VPC Link
EB Environment
Microservice A
Amazon
EC2
Proxy
EB Environment
Microservice B
EB Environment
Microservice C

AWS
API Gateway
AWS
CloudFront
AWS
Route 53
AWS
Lambda
AWS
DynamoDB
Network Load
Balancer
Amazon
Route 53
VPC Link
Amazon
EC2
EB Environment
Microservice A
Proxy

AWS
API Gateway
AWS
CloudFront
AWS
Route 53
AWS
Lambda
AWS
DynamoDB
Network Load
Balancer
Amazon
Route 53
VPC Link
Amazon
EC2
EB Environment
Microservice A
Proxy
EB Environment
Microservice A
Weighted
Routing
99%
1%

AWS
API Gateway
AWS
CloudFront
AWS
Route 53
AWS
Lambda
AWS
DynamoDB
Network Load
Balancer
Amazon
Route 53
VPC Link
Amazon
EC2
Proxy
EB Environment
Microservice A
Weighted
Routing
100%
EB Environment
Microservice A

AWS
API Gateway
AWS
CloudFront
AWS
Route 53
AWS
Lambda
AWS
DynamoDB
Network Load
Balancer
Amazon
Route 53
VPC Link
Amazon
EC2
Proxy
Weighted
Routing
EB Environment
Microservice A
99%
Other tech stack
Microservice A
1%

AWS
API Gateway
AWS
CloudFront
AWS
Route 53
AWS
Lambda
AWS
DynamoDB
Network Load
Balancer
Amazon
Route 53
VPC Link
Amazon
EC2
Proxy
Weighted
Routing
Other tech stack
Microservice A
100%

THE AUTOMATED CI/CD-PIPELINE ALLOWS US TO
SAFELYDEPLOYNEW LAMBDA FUNCTIONS Git
- Lambda code
- AWS CloudFormation
- Jenkinsfile
git
AWS
CodeDeploy
AWS
CloudFormation
Amazon
S3
Jenkins
Build and
upload
Stack
create/update
AWS CloudFormation
- Creates new Lambda version
- Creates a new alias
AWS CodeDeploy:
- Canary deployment (e.g. 10/90)
- Rollback with Amazon
CloudWatch Metric
Amazon
CloudWatch
10%
Code V2
Lambda
90%
Code V1

DATA IMPORTS NEED TO BE
COORDINATED WORLDWIDE
Microservice A
On-Premises
Data sources
On-Premises
Data sources
Microservice B
Microservice D
import
import
import
Microservice C
aggregate
aggregate
10:20 pm
10:30 pm
10:40 pm
10:50 pm
11:00 pm
11:10 pm
05:20 am
05:30 am
05:40 am
05:50 am
06:00 am
06:10 am
06:20 am
UTC+0
A
B
D
C
A
B
D
C
A
B
D
C
A
B
D
C
A
B
D
C
A
B
D
C
A
B
D
C
A
B
D
C
A
B
D
C
A
B
D
C
A
B
D
C
A
B
D
C
Eventual
consistency
Germany Austria France Spain
USA Canada Mexico
...
...
UTC+1
UTC-6
Data processing per day

HOW WE CONTINUOUSLYIMPORT DATA FROM ON-PREMISES DATA SOURCES
Amazon
ECS
ment
bnet A/B
zon
C2
ing Group
Amazon RDS
(Multi AZ)
DB Subnet A/B
Security Group
y Group
Microservice A CloudWatch Event
AWS
Lambda
Trigger
Lambda
AWS Batch
Queue
Tasks
On-Premises
Data
Sources
Execute Task and
Import Data
Create
Job with
Dependencies
A
B
C

HOW WE CONTINUOUSLYIMPORT DATA FROM ON-PREMISES DATA SOURCES
Amazon
ECS
ment
bnet A/B
zon
C2
ing Group
Amazon RDS
(Multi AZ)
DB Subnet A/B
Security Group
y Group
Microservice A CloudWatch Event
AWS
Lambda
Trigger
Lambda
AWS Batch
Queue
Tasks
Create
Job with
Dependencies
A
B
C
Amazon ECR

LEVERAGE AWS LAMBDA TO PERFORM THOUSANDS
OF VEHICLES CONSTRUCTABILITYCHECKS Use case
• Purpose-driven calculation
• Fast response
AWS Lambda
Orchestration
Amazon S3
Vehicle product data
and constructability
rules
Request
Response
AWS Lambda
Constructability
Check
AWS Lambda
Constructability
Check
AWS Lambda
Constructability
Check
.
.
.
Solution
• Spawn parallel Lambdas
Limits
• Restricted parallel
executions
• Reusage of hot lambdas
• Caching prevents
recalculation
• Consider the costs

ON-DEMAND DEVELOPMENT ENVIRONMENTS ARE AN
ESSENTIAL PART OF OUR AUTOMATED TEST PROCESS
Jenkinsgit
AWS
CloudFormation
Stack(s)
On-demand stack for feature branch
Key Facts:
• Gitflow workflow
• Full stack “on demand”
Benefits
• Teams are not blocked
• Isolated testing
• Verification of AWS
CloudFormation code
• Happy Dev teams
Trigger
create
stacks
deploy
feature branch
Microservice with
feature branchAmazon API
Gateway
Amazon
Route 53
Network Load
Balancer
Amazon
Route 53

WE INTRODUCED GAMEDAYS TO TRAIN OUR OPERATION TEAM TO BE WELL
PREPARED FOR POTENTIAL PRODUCTION PROBLEMS
§ Solve real production scenarios
§ Gamification is used through scoring and
leaderboards
§ Chaos Monkey Team brings fun into the situation
§ Each team gains points for
§ Availability of the system
§ Proposed improvements for DevOps-Cycle

FEEDBACK AFTER THE FIRST GAMEDAY
The gameday showed that it’s a great opportunity
to deepen the transferred knowledge in a fun way
and to gain hands on experience through
simulated incidents for the operation work.
“
Operationsteam, March 2018
„

LESSONS LEARNED FROM THE TRANSITION TO AWS
• Don’t build a distributed monolithic application
• Don’t try to make it too perfect
• Stop talking, start building
• Continuous learning and adaptation
• Be aware of AWS soft & hard limits, and build your architecture accordingly
• Doing the transformation in-house was a game changer
• Consider AWS Professional Services as a coach at the beginning of your transition

Best practices

Start with domain-driven design
• Focus on the core domain
• Put a domain model at the center
• Collaborate between
technical and domain experts
Image: Eric Evans, CC-BY-25

Adopt microservices principles
Do one
thing wellIndependent
Decentralized
Black box
Polyglot
You build it, you run it

Adopt the three ways of DevOps
1. Systems thinking
2. Amplify feedback loops
3. Culture of continual
experimentation and
learning

Simple serverless microservice with AWS Lambda
Static
content
User interface Microservices Data store

Simple serverless microservice with containers
Static
content
User interface Microservices Data store

Simple messaging between microservices
Message
Bus
Important
event
Important
event
Provider
Microservice A
SNS Topic A Message
queue
Consumer
Microservice B

app-dev3à 10.1.0.8
app1-tst à 10.1.0.11
db1-tst à 10.1.0.14
app2 à 10.1.0.16
db2 à 10.1.0.18
my-app à 10.1.0.20
websrv1à 10.1.0.1
websrv2à 10.1.0.2
websrv3à 10.1.0.4
app-dev1à 10.1.0.9
app-dev2à 10.1.0.5
db-devà 10.1.0.19
Simple service discovery using Amazon Route 53
app-dev3à 10.1.0.8

Wrap-up/summary
Monoliths aren’t necessarily bad ― in the beginning
Moving to microservices has a lot of advantages
Become a Builder: learn, build, experiment, iterate
Lots of best practices available
Domain-driven design, microservices, DevOps
AWS managed services help make stuff easier
AWS Lambda, Amazon API Gateway, Amazon ECS, Amazon Route 53, and friends
Work with AWS Professional Services, AWS Support, and AWS partners

Thank you!

Building BMW Group's Customer Engagement Platform on AWS (AMT305) - AWS re:Invent 2018

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to Building BMW Group's Customer Engagement Platform on AWS (AMT305) - AWS re:Invent 2018

Similar to Building BMW Group's Customer Engagement Platform on AWS (AMT305) - AWS re:Invent 2018 (20)

More from Amazon Web Services

More from Amazon Web Services (20)

Building BMW Group's Customer Engagement Platform on AWS (AMT305) - AWS re:Invent 2018