8. Analyse user behaviour in Mobile
Advertising domain
User Activity
timestamp
user-id
location-id
device-id
device-
orientation
served-ad-id
clicks
downloads
revenue
User
user-id
age
gender
interests
Location
location-id
city
country
Device
device-id
manufacturer
os
model
9. Operational Analytics
User Activity (by demog, geo, device)
Click activity of users by city and age
Download activity by gender and country for iOS/
Android
Exploratory Analytics
Download conversion by device orientation
(landscape/portrait)
15. Complexity is a silent
killer!
Data Inconsistencies
High engineering and operation cost
Moving data across systems is non trivial
Confusion among users
Multiple definitions of data
Different way of access
Data Silos - Data Discovery
16. What is desired ?
Easy and Consistent mechanism to
discover and query all data
Cost and performance trade-off knobs
for different queries
21. Traditional DWH
Exploratory
Frequently used Data
Low latency response
Fresh Data
Low latency response
Realtime store
All Data
High latency response
Batch store
Batch
ETL
Streaming
Aggregations
Batch
ETL
Unified
View
LENS
CUBE
Operational
23. Lens Capabilities
OLAP Cube Abstraction
Data Discovery via single metadata layer
Query Life Cycle Management
Data Optimisation via Query Analytics
Fast Workload based experimentation
with newer systems: Spark, Tez, AWS
Redshift etc.