A mobile telecommunications company operating in 20+ countries with hundreds of millions of customers.
The solution was to be deployed to all operations, starting with the second largest (60+million customers).
The Business Challenge
The objective was to support hundreds of concurrent random-access queries for customer profile and call data (recharge, voice, data calls) at low-latency.
Customer data was available in the batch-based EDW but this could not support operational queries at scale.
Customer data should be available to downstream systems such as the Call Centre, mobile app, and other third party systems as close to real-time as possible.
Design, install and configure a distributed, highly available, high performance database to serve hundreds of concurrent, random-access queries per second.
Develop an in-memory, paralleled micro-batch ETL to ingest and validate EDW data, and insert into the low-latency database with minimal delay.
The low-latency database data was exposed via secure REST API to third parties such as the Call Centre app.
The database and ETL supported 2+ billion rows/day (catering for history and peak normal throughput), stored for 100 days, with random retrieval latency of less than 1 second.
Both the database and the ETL could be scaled horizontally to match natural data growth.
The Value Proposition
Future use cases such as real-time mobile app, self-service, and eventually realtime fraud, targeted marketing and other analytics would also require this functionality. Therefore the telco’s digital transformation strategy included allowing low-latency, highly concurrent access to customer call and profile data.
The more well-established EDW would have been expensive to scale, and so Big Data technologies and commodity hardware would provide this service at a greatly reduced cost. The Big Data platform would grow to provide other BI and operational services.
No posts were found.