The Coffee Bean & Tea Leaf(CBTL), Business strategy case study
Aon Cloud
1. Prepared by Aon Inpoint | February 2016 | Data Classification: NBI
Cloud Automation & Architecture for Analytics
Aon Centre for Innovation & Analytics
Karl Heery
AWS Summit – London 2016
2. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 2
140+ staff
Data Analysts | Data Scientists
| Business Analysts | IT Development,
Database & Infrastructure Specialists
Platforms, Projects & Services
multi-channel web portals | ad-hoc
reporting | statistical analysis |
machine learning initiatives
Dublin Centre for Innovation and Analytics at the heart of Aon Inpoint
Agile Scrum & Kanban
18 cross-functional teams |
2 weekly sprints |
incremental releases
Aon Inpoint & ACIA (Dublin)
3. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 3
One of the world’s largest repositories of risk
and insurance placement information
Aon’s Global Risk Insight Platform (GRIP)
Telling stories with data
Global Data Pipelines | Change Data Capture | Structured/
Semi-structured | Spikey Workloads
Global Reach | User Experience (UX) focus |
Visually Impactful
4. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 4
Typical reference architecture for analytics
5. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 5
Drivers for AWS cloud adoption
Performance and Productivity
Poor server performance
Re-purposing/refreshing hardware
Capacity planning fails
Cumbersome work practices
Engagement
Believe in better
Focus on business differentiation
Promote experimentation & fail-fast
Drive innovation
Develop careers
Costs and Risks
Poor utilisation
Responsiveness to change
Emerging security standards
Ageing hardware / EoL
Separation of duties
Platform for Growth
Global user base
Data increase across 4V’s
Auto-scaling analytics
Democratisation of data
Relentless business appetite
6. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 6
In order to carry a positive action…
…we start with a positive vision:
2015 VISION
To deliver a secure infrastructure
environment for our Analytics,
with the compute power to match
our appetite
Productionisation
Risk/View Build
Path-to-Live
Compliance & Security
Experimentation
Comms & Education
7. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 7
ACIA – AWS cloud transformation roadmap (2015)
8. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 8
Maturing our Cloud Engineering discipline
• Cross-functional long-living Agile team
• 2 week sprint cycles in cadence with ACIA product teams
• Build dozens of servers & PaaS | Run a fleet of 100’s
• Process-controlled, documented & standardised
• Daily stand-ups; retros, show & tells etc.
➔ Strong focus on self-optimisation – process & architecture
…the journey continues:
2016 VISION
To be architecturally-driven and
automation-focussed in our delivery
of resilient, scalable and cost-
effective Analytics
ACIA Agile
Cloud Team
Global Aon Tech
Teams
ACIA Agile
Delivery Teams
9. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 9
Guiding Principles
✓ Watch utilisation - don’t run what we don’t need
✓ Immutable only - no more pet servers!
✓ Think elastic/ephemeral at design time
✓ Seek out opportunities for serverless
✓ Design for failure so that nothing fails
✓ Automate cloud deployment like any other pipeline
POWERED BY
Continuous Improvement Initiatives 2016
Tune instances based
on CloudWatch
RESTful APIs thru API
Gateway, Lambda &
Swagger
Orchestration
through Docker,
Ansible/Puppet
Deploy SimianArmy
by NetFlix
(in our playground)
Auto-scaling ETL with
S3 & ASG
Open Source for
elastic workloads
Deploy to a parallel
prod stack
CloudFormation repo
for all provisioning
Assess Reserved
Instance needs; scale
for the rest
Tag-driven
Auto-shutdown
10. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 10
Everything’s in containers these days!
AWS Re:Invent 2015 – the after-party venue
11. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 11
Data Science Containerization through Docker
▪ Begin with checkout of a base template
▪ Docker-compose configures application services
▪ Developer maintains dockerfile for env configuration
▪ Everything checked in together
➔ Code wrapped with dependent runtime, tools &
libraries, for full portability
12. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 12
Docker Deployment Pipeline using Amazon ECS
▪ CI server handles
promotion of container
to repository manager
▪ Deployment server
pushed new task defn.
to ECS
▪ ECS refreshes and
redeploys container to
cluster of instances
▪ Airflow preferred for
scheduling and wider
orchestration
13. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 13
Final deployment view – Analytics Engine
14. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 14
• Lean in - Sleeves-up - Hands-on
• Culture of experimentation
• 15 certifications in 8 months
• Innovator’s duty to share across Aon
Building competency, changing culture…
“The only source
of knowledge is
experience”
15. Aon Inpoint | February 2016 | Proprietary & Confidential | Data Classification: NBI 15
Contacts
Karl Heery
Head of IT
Aon Centre for Innovation & Analytics (Dublin)
karl.heery@aon.ie