The Trial Company is the largest supermarket chain in the Kyushu region of Japan, selling a range of discount food, apparel and electrical goods. Trial operates 148 branches in Japan, five branches in South Korea and a logistics and software research and development center in China.
With US$1.7 billion in annual revenues in 2011, Trial is focused on expanding its distribution and retail business across Japan – and further into Korea and China – with a goal of having 800 retail outlets within five years. This means the company must develop its distribution and retail systems in parallel with its distribution network to enable a self-sufficient supply chain.
Scaling Data Analysis Capability to Meet Massive Growth
Prior to deploying Pivotal Greenplum, Trial built its own data warehouse and business intelligence (BI) system and chose to run a third-party relational database management system (RDBMS) on the database. But the system could not scale sufficiently to meet the company’s increasing volumes of transactional data, resulting in a decrease in data processing speed. Unable to analyze data in real time to meet specific business requirements, the supermarket chain struggled to remain competitive.
“Our transactions grew to more than 500 million per month,” says Go Murano, Section Chief, Data Systems Division, Management Head Office, The Trial Company. “Our existing system couldn’t meet our storage demands because, even as we added disk space, we couldn’t find a way to increase our data analysis and processing speed.”
Pivotal Greenplum Supports Expansion
To meet its analytical processing and data management requirements, Trial chose Pivotal Greenplum, which is built on an open-platform, “shared nothing,” massively parallel processing (MPP) architecture. Trial deployed Pivotal Greenplum alongside its scaled-out proprietary system, which uses a Unicage development approach based on OS commands and shell scripts.
Trial needs to manage large volumes of unstructured data — such as image files. To that end, the company has been testing Apache Hadoop for high-level analysis of unstructured data. Pivotal Greenplum supports reading files from and writing files to the Hadoop File System. Pivotal Greenplum also supports MapReduce programs and provides Trial with highscale data analysis capabilities by enabling programmers to run analytics against datasets — whether they’re stored in the Pivotal Greenplum or elsewhere.
Reduced Data Analysis Costs
Prior to Pivotal Greenplum, Trial had considered proprietary database management systems that required dedicated hardware, which would have driven up project costs, caused development delays and further hindered Trial’s expanding distribution network. By contrast, Pivotal Greenplum uses commodity servers, storage and Ethernet switches and supports industry standard interfaces. This meant Trial could cost-effectively link its existing BI applications and reporting tools to the new system.
“Other solutions required expensive, dedicated hardware for which we would have had to modify and possibly re-build our applications,” recalls Masayasu Fukushima, Deputy Director, Data Systems Division, Management Head Office, Trial. “Our testing showed that for the same amount of storage, migrating to a system running Pivotal Greenplum would cost 50 percent less than other systems running a MPP solution.”
Trial’s new data warehouse solution, based on Pivotal Greenplum, ensures approximately 6 terabytes of effective storage volume. It is designed to allow Trial to add segment servers as required, expanding its processing and storage volume as it opens new retail outlets over the next five years. Because every primary segment instance has a corresponding mirror segment instance, Trial can count on high availability during server failure. The database will stay online in the event that the mirror segment ever takes over from an unavailable primary segment. This functionality will be critical in helping Trial support its distribution and logistical operations and maintain supply to its retail outlets.
Faster Processing Speeds
Since going live on Pivotal Greenplum, Trial has significantly improved its data analysis task run times, with processing speeds increasing by up to 250 times for some applications. These faster results are driving more efficient decision-making — even on the mobile handset PACER, which Trial uses as an information terminal, workstation and phone. Employees who are tasked with accessing specific data have improved their mobility by retrieving the information directly from the handsets at five times the speed of using client PCs.
Reduced Space Requirements
Due to in-database compression with the highly efficient Pivotal Greenplum, Trial has not only increased performance, but also dramatically reduced the space it needs to store data. The company has achieved a 70 percent disk space reduction, taking 2 terabytes of data down to 600 gigabytes, after migrating to Pivotal Greenplum.
With Pivotal Greenplum, Trial is now conducting an extensive range of complex analytical tasks that were previously impossible for the company. “Before Pivotal Greenplum, we couldn’t undertake accurate tracking of popular products by region because our old solution couldn’t produce a complete customer distribution chart,” says Murano.
Trial now has access to real-time competitive data that is improving strategic decision-making and supporting the company’s move toward complete self-sufficiency within its supply chain. “By making use of cutting-edge IT systems, like Pivotal Greenplum, we’re striving to achieve in one year the type of change it takes a conventional company 10 years to achieve,” concludes Hiroshi Yamanaka, Division Director, Data Systems Division, Management Head Office.