Call Detail Record (CDR) Analytics Platform

แพลตฟอร์มวิเคราะห์ข้อมูล Call Detail Record

Collects CDRs from multiple networks into a 580 TB Hadoop data lake, processes them into data layers with Hive and Impala, and delivers web reports that operations teams use directly.

Overview of the CDR analytics platform and its 580 TB data lake

About the project

Puts Call Detail Records (CDRs) to work for a telecom operator's operations teams.

Data path

Ingestion — CDRs from SGSN, PGW and MSC across several networks enter a single pipeline, covering voice, SMS, USSD and supplementary calls
Processing — a group of Miner (ETL) nodes extracts, converts and stores the data as RAW, ORC, L1, L2 and report layers, with raw backups
Storage — a 580 TB data lake on Cloudera (CDH) / Hadoop
Aggregation — Hive and Impala summarise the data so web reports load quickly

Web reports

Monthly usage, split into upload and download
A Transform screen that tracks processing status
3G/4G traffic by region with maps, cause-for-record-closing breakdowns and the top 10 abnormal releases
MVNO view: voice and data usage share per provider, with regional capacity tables
Monthly data roaming performance by province and closing cause

Technologies

Hadoop / CDH · Hive · Impala · Spark · ETL Miner · MongoDB · Web Report

Gallery · 4 images

Architecture from CDR through ETL and the data lake to web reports

Monthly usage report and processing status screens

Regional traffic and record-closing cause reports