Telecom Bell is migrating their core applications to the cloud to improve network quality of service and enable personalized customer engagement using customer data. They are facing challenges with their on-premise data platform's lack of scalability, data silos, and governance issues. Databricks will help design a new cloud-based data platform architecture using their platform and Confluent for event streaming. The joint delivery approach between Telecom Bell and Databricks teams will include establishing data governance, migrating applications in phases, change management support, and reaching the desired timeline of May 2024.
3. Summary | Business challenges
1
1 2 3 4 5
TELECOM BELL must
improve network
QOS to align with
consumers' changing
emphasis on mobile
connectivity and
data usage
As IoT and 5G advance,
customers easily switch
providers, prompting
TELECOM BELL to
prioritize personalized
engagement using
customer data for
customized messaging
and services.
TELECOM BELL is subject
to many regulations,
including data privacy
and security regulations,
and needs effective ways
to adhere to these.
Power of data
there is a data-volume
explosion, requiring both
focus and new
capabilities.
Increase pressure to show
growth and profits
is constant and data and
AI will be a critical
enabler
4. Summary| Technical Challenges
2
Today there are increased expectations and pressure on the Telecom organization to have a strong data & analytics strategy
•Data platform is not scalable for analytics, AI/ML
Upfront capacity planning and cost
Governance of the data on HDFS is a challenge
Data sits in silos and not easy to integrate/ connect
•Lack of discoverability of data (catalog)
•Housekeeping - Maintenance of the in-house cluster is a difficult thru
different portals and installations
•Advance disaster recovery, durability and availability
•Bigger IT infra staff required
5. Summary | Executive Plan
Telecom Bell wants to improve the Quality of Service (QoS) of their network and to get there, start
migrating the core applications to cloud.
Databricks will bring industry leading expertise and Databricks platform expertise to drive the
transformation at speed.
Confluent will bring event streaming platform built on Kafka and the necessary platform support
Telecom Bell has a team of 10 Engineers with expertise on Kafka and spark
Desired timeline – May 2024
4
1
3
2
5
3
7. Platform & Architecture | Current Architecture
1
Limitations
• Data platform is not scalable for analytics,
AI/ML
• Upfront capacity planning and cost
• Governance of the data on HDFS is a
challenge
• Data sits in silos and not easy to integrate/
connect
• Lack of discoverability of data (catalog)
• Housekeeping - Maintenance of the in-
house cluster is a difficult thru different
portals and installations
• Advance disaster recovery, durability and
availability
• Bigger IT infra staff required
8. Platform & Architecture | End state Architecture
2
Design target state architecture for a scalable, secure and well governed data platform
(AI /ML self-serve, advanced engineering capabilities including necessary governance on lake capability)
Highlights
• Warehouse + Data Lake capabilities at scale with
Governance
• Data product mindset – Marketplace, Self service capabilities
• MLOps – Full ML Lifecycle
• Domain data tiers - Advance data management capabilities,
curated democratized data layers
Designing and activating a World Class Data
Platform:
Fundamental Principles
• Scalability
• Performance
• Industrialized processes governing the pipeline
• Distributed, fault tolerant architecture
• Open file format for better interoperability between systems
• Security and reliability
• Data provenance and lineage
• ACID complaint
9. Platform & Architecture | Current vs New
3
More performant and optimized spark
engine
1 Governance under the same roof
2
New
10. Platform & Architecture | Artifacts
4
Key components of the data platform:
A World Class Data Platform!
12. Approach | Our Tenets
1
Security is job
zero
Agile
Methodology
Continues
delivery of
results
Because - "Approach is the first step towards achieving goals"
Leverage customer
asset first
Multiple velocity
joint delivery
approach
A B C D
E F G H
Zero down time Log the journey at
every step to look back
& learn
Principal of least
access
privilege(PoLAP)
13. Approach | Objectives
2
Build the data strategy
roadmap that
empowers Telecom Bell
to overcome its
business challenges
Mindset
HORIZO
N
HORIZO
N
HORIZO
N
Strategic
roadmap
Platform
1
2
3
Build strong
foundations with data
platform development
and implementation
Co-create an operating
model that would take
TELECOM BELL where it
wants, in a sustainable
way.
Migrate core
applications to cloud in
a secure and reliable
way
4
Industrialization
15. Operating Model | Joint Delivery Approach
Executive Leadership
Databricks Leadership:
1
Application Team
Telecom Bell Leadership
1
Program Management
Databricks Lead
1
Telecom Bell Lead
1
Platform Team Data Quality &
Governance
Bringing it Together
Databricks
(Professional services)
5
D
C
B
A
Meeting Cadence
• Bi-Weekly Steering
Committee Meetings
• Weekly PMO Meetings
• Daily Delivery Team
Meetings
Telecom Bell Resources
3
1
Telecom Bell Resources
3
Telecom Bell Resources
4
Telecom Bell Resources
1
Databricks
(Professional services)
5
Databricks
(Professional services)
3
Databricks
(Professional services)
3
16. Leadership
Scrum Master
Application
Team
Functional
Domain Expert
Data Visualization
Engineer
Customer Success
Engineer
Data Engineer
Operating Model | Pod Structure
Data Quality &
Governance
Test /
Quality Lead
Data Quality
Engineer
Data Governance
Lead
Data Lineage and
Profiling Engineer
Product
Owner
Bring it
Together
Delivery Lead
Change
management
Specialist
PMO Lead
Roadmap
Officer
Databricks resource Telecom Bell resource
Leader
Leader
Leader
Leader
Platform
Azure Platform
Cloud Architect
Cloud DevOps
Engineer
Resident Solutions
Architect
Delivery Solutions
Architect
Customer Success
Engineer
Resident
Solutions
Architect
Resident
Solutions
Architect
Specialist Solutions
Architect (Security)
Specialist Solutions
Architect (Security)
Cloud DevOps
Engineer
Scrum Master
16 12
Shared
Resource
Shared
Resource
Shared Resource
Cloud DevOps
Engineer
2
Enterprise
Support
Enterprise
Support
17. CELEBRATION
Celebrate completion
`
PROGRAM
KICKOFF
Operating Model | Road Map
3
DELIVERABL
ES
DIAGNOSTIC OF THE CURRENT
ENVIRONMENT
1
PLATFOR
M
3
END STATE ARCHITECTURE
2
MIGRATION: 10
%
4
MIGRATION: 60%
5
6 MIGRATIO
N
100%
Progress
Progress
Consistently –
communicate,
remove
roadblocks &
eliminate
friction
Celebrate
completion of
quick wins to
strengthen
morale
ALONG THE WAY
Progress
MEASURE PROGRESS
MIGRATION
PLAYBOOK
A repeatable guideline to
migrate
applications to new
architecture
3
HUMAN-CENTERED
CHANGE
Focus on each individual team
member’s technical skills and
capacity for change. Reskill team
members whose roles are changing
1
MINDSET CHANGE
Adopt ‘Data as a Product’, self
service platform, federated
governance, domain specific
ownership
2
PROCESS GOALS
18. Operating Model | Timeline
3
Q2 2023 Q3 2023 Q4 2023 Q1 2024 Q2 2024
Agile : Update Roadmap and plan per evolving
priorities
Current State
Diagnostics
Assess skill and capability
gaps within the organization
Design & Deliver Governance Structure
Databricks workspace
setup
Assess Current State & Catalog Critical
Data Elements
Prepare Governance Strategy
(Identify roles, define interaction model)
Application
Platform
Bring it together
Data Quality +
Governance
Best practices and tagging
Design Target State DQ Monitoring
Steerco
Meeting
Assess Current State Data Governance
Steerco
Meeting
Steerco
Meeting
Steerco
Meeting
Confluent workspace
setup
Cost management reports
Define
Elements/Sources/Dat
a
Test & Modify
Refactor the
code
Deploy
Document &
KT
Define Pods and
teams
Create Upskilling Curriculum
and setup trainings sessions
Establish ways of working –
documentation, win celebrations
Continuously monitor, foresee risk, mitigate risks , fetch leadership
guidance
Project management
Arrange handover of all
areas
Handover
Handover
Handover
Security and compliance | phase1
Security and compliance | phase2
Talk to business
team
Incorporate changes
Cost optimization
Move towards Infra as
code
Implement Target State DQ
Monitoring
20. Industrialization:
Competitive Differentiation
High throughput of innovation analytics (AI/ML)
Predictive analytics at scale
Data driven(real time what-if analysis)
Harmonized MDM; ML & AI based DQ
Fast, repeatable time-to-market from idea to
product
5
Additional Details | Future Scope
1
21. Additional Details | Risk & Mitigation - Technical
1
Risks Mitigating Actions
Data Loss Risk
Reconciliation, Check pointing, Audit, Monitoring. Use of fault tolerant ingestion/migration tools like Azure Data Factory
– Az Copy Activity
Data Corruption and Data Integrity
Risk
Data Validation - Each record is compared in a bidirectional manner, and each record in the old system is compared
against the target system and the target system against the old system
Interference Risks
(simultaneously use of source
application)
Align with the stakeholders of each source on how the bandwidth can be shared. “Bring it together” team come into
play to address this
Schema Evolution
(Changing Dimensions)
Delta file format – Schema evolution feature. Depends on schema on read. Further to make sure there are no
incompatible schemas coming in. A catalog and governance would be leveraged – Databricks Unity Catalog
Authorization Risk MFA and Identity Federation , access controls at row and column level by Delta Lake
Data Security Risk
Apply Encryption where possible and appropriate
All tokens and keys will be securely stored and rotated in Azure Key Vault
Rotate keys on regular interval
Down time due to migration Replicate and activate approach
22. Additional Details | Risk & Mitigation - Other
2
Risk Mitigating Actions
Resource Availability &
Competing Priorities
Making sure employees are fully advised about participation into workshops and/or interviews.
Get the right people at the right time
Senior Leadership Buy-In and Delays
in Decision Making
Strong support from the leadership Group, including areas who are not fully involved by the initial changes. One Team,
One direction
Establish governance to provide clarity on accountabilities for decision making
Potential Impacts to Other
Projects
Strong support from Senior Leadership if there is a need to put a hold on
existing projects
Review current state of ongoing projects to see how it impacts to the Finance model
Prioritize major changes and focus on the big obstacles upfront
Lack of People Adoption –
Major Change
Agile and inspirational change management and communication structure
Leverage Bring it together team, and roles like change management experts to steward people readiness and prepare
for change
Design in Isolation
(Enterprise Integration)
Work with scalable and flexible design principles in mind to ensure proper
integration and alignment with the business. It is a partnership approach
Gather key inputs to support cross function process design decisions
where applicable
Availability of Key Data Inputs
and Information
Simplify data requests to collect data and information at the appropriate level of detail
Assign designated Databricks and Telecom Bell contact to ensure smooth and timely transition of data
Discovery Phase to identify hidden environmental risks to foresee and mitigate
23. Area Assumption
1 Platform
Telecom Bell on premise platform is owned and managed by Telecom Bell and Databricks will get the necessary support to extent the setup to
provision the solution per the scope of this effort.
2 Data Security
Telecom Bell is responsible for the design, integration and operation of all Client Identity and Access Management, Security Incident and Event
Management, Vulnerability Scanning and Security Testing tooling and processes as appropriate.
5 Access & Setup
Telecom Bell will provide system access to all source systems or applications required by scope. Telecom Bell will provide access to systems and
environments(including DEV, SIT) within 5 business days of receipt of request.
6 Access & Setup
Databricks persona will not have access to unencrypted PII data. Telecom Bell will be responsible for encrypting any PII data, prior to extraction in
the Databricks platform.
7 Access & Setup PII and GDPR Data handling will be done by Telecom Bell as per the existing practices in delivery , any additional arrangement is out of scope.
9
Project
Management
Telecom Bell will provide relevant functional, technical and process documentation for data platforms and systems required by the scope.
10
Project
Management
Telecom Bell will nominate full time business and technical SMEs aligned to this project as per the agreed pod structure.
11
Project
Management
Telecom Bell data owners /nominees will make every attempt to attend the Scrum meetings and ceremonies to present their progress on the issues
assigned
12
Project
Management
Telecom Bell will make sure we get required time and support from all the stakeholders for complete success of the project.
14 Data Build
Databricks team will reuse and extend the existing data ingestion tooling and framework to support the ingestion activities into the platform. The
project will carry a data discovery exercise where it will assess the local market data quality and readiness.
15 Data Build Source System inventory have already been identified and already in place.
16 License The Cloudera CDH on premise license is already expired in March 2022. However, the extended support is required and obtained.
Additional Details | Assumptions
3
24. •Is there an onboarding guide for the consultants to get started on your environment ?
Is there a Source System inventory already identified and can be shared ?
What are the roles and skills of existing 10 engineers on the team ?
What is the current data governance mechanism ?
Other than Cloudera, what all other paid subscriptions and packages are installed on the concerned
architecture ?
•Is there any major business contingency on this project plan? If so, what is the impact of the delayed delivery?
•What are all the compliances and regulations that Telecom Bell need to follow about the concerned data?
•Does Telecom Bell already have Azure account? If so, what is the level of enterprise support plan that is subscribed ?
•Does Telecom Bell already have Confluent account? If so, what is the level of enterprise support plan that is subscribed
?
•Any due license expires ?
•What is the Cloudera’s extended support expiry date ?
Additional Details | Questions
4
26. Yashodhan Kale
BACKGROUND SELECTED EXPERIENCES
Amazon Web Services Certified Data Analytics - Specialty
Amazon Web Services Solutions Architect - Associate
Cloudera Certified Developer for Apache Hadoop (CCDH)
RELEVANT FUNCTIONAL AND INDUSTRY EXPERIENCE
Modern Technologist | Data and ML at scale
Design and drive clients' Data and AI journeys powered by cloud analytics
expertise! Offering data product mindset-driven solutions to deliver platforms
and beyond: Self-service framework, rapid experimentation lab, democratized
data, data products marketplace, multi-cloud solutions, data lake, data fabric,
data mesh patterns with federated governance, domain-specific ownership, and
more
Industry Focus:
• HealthCare
• Retail
• Market Research
• Finance
Functional Expertise:
• Digital Transformation
• Analytics and CDO Strategy
• Open Source
• Machine Learning, IOT
• Data Drive Re-invention
• Fortune 5 American healthcare company
Establish and manage DevOps, Data Engineering, and ML engineering teams in close collaboration with Data
Scientists. Set up a self-service Data and ML platform on Azure cloud for a Retail enterprise, incorporating an
experimentation framework, Model Training pipelines, and real-time inference using Azure AKS, Kubeflow, and
Snowflake. Implement an Rx enterprise Data and ML platform on Azure cloud, enabling ETL pipelines with
Databricks and Apache Airflow. Lead the development of large-scale projects, including legacy modernization,
Rx personalization, and Retail personalization programs that impact millions of lives daily. Collaborate with
technology partners, MSFT and NVIDIA, to present objectives, findings, and incorporate feedback for ML
solutions with specialized NVIDIA GPUs. Architect and oversee the implementation of the Refrigerator IoT
project on Azure, leveraging IOT hub, Azure Analytics, and Databricks. Lead the development of SAP HANA to
Spark integration. Manage the enhancement team in Data Engineering for pharmacy-related projects, ensuring
critical business deliveries. Design data-driven solutions, including self-service analytics platforms, rapid
experimentation labs, democratized data, multi-cloud solutions, data fabric, data mesh patterns with federated
governance, and domain-specific ownership. Develop an ingestion framework for seamless data migration across
projects and cloud storage services.
• Multinational American information, data & market measurement company
Build a retail store data aggregation engine (Retail Intelligence system) for 24 countries, initially using Hadoop
MapReduce, later upgraded to Spark. Migrate on-premise batch processes to the cloud using Docker, Azure Batch
Services, and Azure Shipyard for cost efficiency. Perform performance tuning on Apache Spark, cloud Hadoop
clusters (HDI), and Databricks on Azure and Hadoop platforms.
CERTIFICATIONS
PREVIOUSLY
Sr Cloud Solution Architect @ Amazon Web Services Level 6
Sr ML Engineering Manager @ Databricks Level 6
WHAT HAS BROUGHT ME HERE
• Customer Obsession
• Deliver Results
• Earn trust
• Learn and Be Curious
27. ACID Compliant
Time Travel
Data as product
Inter Operability
Self service
experimentation
Scale &
Pay as you go
Lake House Governance
Data Migration
Identity Management,
SSO
Event Streaming
Exactly once
semantics
28. Upfront cost
Not easy to integrate/
connect
Lack of discoverability
Efforts to make data HA &
durable
End of support
Maintenance
29. Platform & Architecture | Artifacts
1
Key components of the data platform:
A World Class Data Platform!
34. Databricks Notebooks
1
Share
insights
Quickly discover new insights with
built-in interactive visualizations,
or leverage libraries such as
Matplotlib and ggplot. Export
results and Notebooks in HTML or
IPYNB format, or build and share
dashboards that always stay up to
date.
3 Production
at scale
Schedule Notebooks to automatically
run machine learning and data
pipelines at scale. Create multistage
pipelines using Databricks Workflows.
Set up alerts and quickly access audit
logs for easy monitoring and
troubleshooting.
2
Work
together
Share Notebooks and work with peers
across teams in multiple languages (R,
Python, SQL and Scala) and libraries of
your choice. Real-time coauthoring,
commenting and automated versioning
simplify collaboration while providing
control.
Notas do Editor
Ex AWS
Ex, Databricks ML Engineering Sr Manager,
I have built data platforms and delivered - campaign management, personalization while touches millinos of lives a day.
Extensively worked into Retail , healthcare, telecom and finance industries and worked into 3 different counties
experienced start up culture. And I know how to deliver results.
Qualities that has brought me here are – Customer ob, Delivering result, earn trust and not giving up on learning.
fifa, chess, Salsa
Transition –
that’s me . With that lets get going
10K ft overview
business, technical , and plan
slightly deeper look into platform
Transition –
what how and when
personalization customer engagement
regulations - data privacy and security
data volumn
show growth and profits.
Top priority : improve QOS
Transition
-Lets look at some technical challenges
To address the B challenges above A Strong data and analytics strategy is
maintain the pace of innovation =
experimentation capability
=
pay as you go is crucial
+ easy access to data
+ SAAS model of services
====
benchmark , Red flags - : kafka and spark architecture which processes nework data
End of support
Transition –
yes we saw the businesses and technical challenges so What’s the plan ? The direction : next slide
we will improve the QoS of the network and start with migration
Databricks
Confluent
telicom bell10 engineers
we plan to complete this project in 12 months
Transition –
Alright. How do we achieve this? I have but together plan and that I will walk you thru.
feedback, suggestions, concerns are all welcome. Craft a final version together.
TB On -premise arch would look more or less like this.
Transition –
Enough time on architecture. Lets take 2 differences and move from here.
1. More optimized more performant.. With less configurations to worry about. : zordering, vaucum, auto optimize feautures.
2. integration with UC
A leap towards data as a product mind set –
Federated goverenance, self service platform, inter – operatability, share within and across organization – notebooks and code. (product mindset.)
Add Marketplace
Talk about few in the interest of time
Security – Azure key vault,
Encryption where possible.
Network setup – No data will flow thru public internet.. Private endpoints will be used.
Principal of least access privilege(PoLAP)
Zero down time:
Replicate and then activate
Leverage customer asset first :
you will see in the next few slides 10 engineers distributed across all project areas
operating model is designed to deliver these objectives over next 12 months
After the essential piece of roadmap and planning
Platform : Not only run existing apps ,
empowers bell to accelerate on pace of innovation
provide solutions beyond the scope of this project: - personalized customer engagement and other business experiments
Also, OM delivers a needed shift in mindset. Think “Data as a product” , create a data-product culture
features of marketplace, federated governance, delta sharing,
Lastly, it deliver pay as you go, secure & low maintenance solution that can handle the immediate need to migrate to cloud “given end of support”
sharing the resource where possible
used all 10 tb enineers
Assuming enterprise support from confluent and azure
total count
Dignostic -
complete picture of where we are,
our pain points,
scope of improvement,
asset
Final End state architecture
Time line activity
Any party has any concerns, we can definitely relook at this and try adjust to make it smoothly achievable
Ex AWS
Ex, Accenture ML Engineering Sr Manager,
I have built data platforms and delivered - campaign management, personalization while touches millinos of lives a day.
Extensively worked into Retail , healthcare, telecom and finance industries and worked into 3 different counties
experienced start up culture. And I know how to deliver results.
Qualities that has brought me here are – Customer ob, Delivering result, earn trust and not giving up on learning.
fifa, chess, Salsa
Transition –
that’s me . With that lets get going
A leap towards data as a product mind set –
Federated governance, self service platform, inter – operability, share within and across organization – notebooks and code. (product mindset.)
Few pain points -
These services all run on premise. upgrades
Limitations
Data platform is not scalable for analytics, AI/ML
Upfront capacity planning and cost
Governance of the data on HDFS is a challenge
Data sits in silos and not easy to integrate/ connect
Lack of discoverability of data (catalog)
Housekeeping - Maintenance of the in-house cluster is a difficult thru different portals and installations
Advance disaster recovery, durability and availability
Bigger IT infra staff required
A leap towards data as a product mind set –
Federated goverenance, self service platform, inter – operatability, share within and across organization – notebooks and code. (product mindset.)
Add Marketplace