|
Orange
Customer Quick Facts
The largest cellular provider in the country
Established: 1999
Subscribers: 2,600,000
Market share: over 30%
Business objectives
Increase profitability
Improve performance
Strategy
|
Preserve investment
|
Preserve vast investment made in the DWH
|
|
Support new business initiatives
|
Content, media ,3G , advertising
|
|
Improve performance
|
Performance became unbearable upgrading the hardware would cost a fortune
|
|
Add scalability
|
Size of data is increasing dramatically and new sources are continuously being added
|
|
Stability recovery capabilities
|
Increase availability for the users , reduce night window , recover from disasters
|
|
Real-time capabilities
|
Have the ability to retrieve Data while loading , Trickle feed
|
Gilon's Solution Objectives
Database – Improve data model, redesign architecture, DB tuning, handle scalability, handle performance.
ETL – Redesign ETL strategy, improve performance, robustness, scalability.
Data Quality – Data profiling, implement DQA procedures, continuous data improvement.
Methodology – Development standards, versions release procedures.
Gilon Implementation
The challenges at the company's DWH performance became critical due to the rapid growth of the company in subscribers' and thus in usage volumes.
The company started with 100,000 subscribers and grew to several million in 2-3 years. Consequently there was a significant increase in ETL processes, more complicated logics were added, the processes became unstable and the entire DWH situation became mission critical.
Gilon's approach was to develop a high level design for all processes in three different channels:
- Build new processes correctly
- Improve existing processes
- Improve access to Oracle
Implementation included the following processes:
Mapping problems with a global view
- Building processes the right way
- Incremental load
- Generic mechanism for identifying deltas
- Generic mechanism for identifying target population
- Splitting complex maps
- Recovery mechanism from last fall
- Reuse of generic objects
Improving performance in Informatica
- Reduced lookups
- Use of sorted aggregates
- Use of joiners with sorted merger
- Use of update strategy
- Limit the population as early as possible
Improving performance in Oracle
- Use of analytical functions whenever possible
- Insist on the correct explain plan
- Remove indexes before full load
- Extensive use of snapshots
- Extensive use of partitions
- Bulk updates
The results
Improved performance, scalability and robustness which allowed the company to add processes according to business needs and significantly improve usability scope of the DWH.
|