CASE STUDY - Turkcell Transforms Its Business with High-Performance Data Integration and Data Warehousing
Commentary by Ferhat Sengonul, Senior DBA, Turkcell
Turkcell is Turkey’s largest GSM carrier, with more than 34 million subscribers. Since Turkcell started its operations in 1994, it has continuously increased the variety of its services based on mobile, audio, and data communication; its quality levels; and as a result, its number of subscribers. Turkcell is also the third biggest GSM operator in Europe in terms of subscriber numbers. Turkcell was founded in 1994 and is headquartered in Istanbul, Turkey.
Turkcell implemented an enterprise data warehouse solution that supports many business decisions, including subscriber segmentation, credit scoring, and churn analysis. Fast access to these critical reports is important to Turkcell’s success in a highly competitive industry. As the number of subscribers increased rapidly over the past decade, Turkcell’s data volumes reached 250 TB of uncompressed data and the company had to add more servers, for a total of 11 server racks.
“Even though we invested in more CPUs for the environment, both query and ETL performances were not adequate to meet the needs of the business users, and architecture complexity led to increased IT operations costs,” said Ferhat Sengonul, senior DBA at Turkcell.
In addition, Turkcell needed to improve its ETL process performance for loading the network operations data warehouse, running on Oracle Database 11g Release 2. The company also wanted to integrate more network operations systems to the data warehouse to offer a global reporting solution for the company.
After a thorough evaluation, Turkcell selected Oracle Exadata for its data warehousing solution due to its high performance, simplified architecture, and linear scalability. In addition to supporting reporting and data mining, Oracle Exadata consolidates the overall database infrastructure. Turkcell chose Oracle Data Integrator (ODI) as its data integration solution to provide bulk data loading into Oracle Exadata. With its extract, load, and transform (ELT) architecture, ODI performs transformations within the data warehouse.
“In addition to providing higher performance and reducing cost of ownership, Oracle Data Integrator and the Oracle Exadata solution allow us to turn data into valuable information that moves our business forward.” —Ferhat Sengonul, Senior DBA, Turkcell
Turkcell also implemented ODI in its network operations data warehouse project, which enables the company to optimize network operations. ODI extracts from systems that run on Oracle 10g, Oracle 11g, Microsoft SQL Server 2005, Microsoft SQL Server 2008, Sybase ASE 15.02, Sybase IQ 12.7, and different types of files produced from OLTP systems.
By deploying Oracle Exadata, Turkcell decreased the number of server racks in its data center from 11 to 1, and went from 250 TB of uncompressed data to 25 TB. “We also achieved, on average, 10 times performance increase in query response times,” said Sengonul. Also, by moving from a traditional ETL architectural approach using Oracle Data Integrator’s ELT architecture within Oracle Exadata, Turkcell improved its ETL processing performance by 500 percent. The combined solution helped the company decrease data preparation time for data mining from 27–29 days to 2–3 days.
With the new and lean data integration architecture, the network operations data warehouse has higher performance. Because of the hourly in-day ETL mechanism, network traffic data can be accessed in the next day. The network operations data warehousing solution enabled Turkcell to reduce network operations costs by decreasing network faults. The company can forecast network alarms, provide faster responses, and improve customer satisfaction. In addition, the company can better assess equipment and materials utilization and purchase requirements. Furthermore, Turkcell can better optimize its network operations by gathering information from the whole network infrastructure using ODI’s fast performance and heterogeneity.
This article originally appeared in the issue of .