CASE STUDY - SoftBank Mobile Improves Database Query Performance by up to Eight Times
Commentary by Keiichiro Shimizu, Senior General Manager, Planning Management Department, Information Systems Division, SoftBank Mobile Corporation
Established in 1986, SoftBank Mobile Corporation is a leading mobile telecommunications service provider based in Tokyo, Japan. It offers a range of mobile services that run on Wideband Code Division Multiple Access (W-CDMA) and Universal Mobile Telecommunications System (UMTS) 3G networks.
Driven by the popularity of smartphones, SoftBank Mobile has achieved the highest growth in Japan’s mobile phone market over the past two years, attracting more than 200,000 subscribers per month. This increase in subscribers from a previous average of 50,000 per month strained the company’s data warehouse.
“We quickly realized that with this expected increase, we would run out of storage capacity by the end of March 2010,” said Keiichiro Shimizu, senior general manager of the planning management department, information systems division, of SoftBank Mobile. “Performance problems started to occur; it was taking us 25 hours to analyze the data log each day.”
Database Performance Increased up to Eight Times
In January 2009, SoftBank Mobile conducted an Oracle Exadata Database Machine proof-of-concept test with a data volume based on an expected future increase in transactions. During this test, the company’s data warehouse performance improved by up to eight times.
As a result, SoftBank Mobile determined it was able to replace 36 Teradata racks with just 3 Oracle Exadata racks. The new data warehouse, running on Oracle Exadata, is connected to the customer care and billing system, which runs on Oracle Database. It can store up to 150 TB of data, an increase in capacity of 150 percent on the previous Teradata solution.
Faster Performance at a Lower Cost
The intelligent storage software in every Oracle Exadata Database Machine enables the company to offload processing from its database server to the storage servers. This provides significant database performance improvements by reducing database server CPU consumption while eliminating network bottlenecks. For example, it now takes only 7 hours to analyze call records and customer logs each day, compared to 25 hours previously. This enables SoftBank Mobile to serve customers faster based on common call detail records and customer logs, which has strengthened the company’s marketing power.
Operational Costs Reduced
SoftBank Mobile’s operational costs have been significantly reduced since the introduction of Oracle Exadata Database Machine.
“Oracle Exadata Database Machine has enabled us to create a data warehouse with up to eight times the processing capacity of our previous data warehouse, while reducing our overall database running costs by 50 percent,” Shimizu said.
The introduction of Oracle Exadata Database Machine has also helped increase staff skills by eliminating proprietary technology and utilizing more than 100 Oracle master engineers. “It’s also much easier to get access to engineers who have the relevant experience working with this system, and our operational costs are less than half of what they were previously,” said Shimizu.
SoftBank Mobile initially considered updating its existing data warehouse, but decided it would be too costly. The company then compared five solutions before selecting Oracle Exadata Database Machine.
According to Yuji Watanabe, deputy manager of the operations department, business base management department, information systems division, of SoftBank Mobile, Oracle Exadata Database Machine is an open and high-performing system. “The technology is also based on Oracle Database, a simple architecture that we understand, so we knew we would succeed implementing the system,” he said.
Masaki Matsuoka, a project manager for the information systems division, business base management department, project promotion department, of SoftBank Mobile, traveled to the U.S. in March 2009 to observe the system at one of Oracle’s test centers.
“The system processed a data volume exceeding one trillion items within a few seconds,” said Matsuoka.
The implementation took just three months after system analysis and base design phase, and was completed in May 2010. The Oracle development and Oracle field teams fully supported this project.
This article originally appeared in the issue of .