For its new information analysis platform, LIXIL selected a combined system of the open source software, Hadoop, and SAP’s in-memory database SAP HANA. Hadoop is used for the processing, basic analysis, statistical analysis, and machine learning of access logs from external SNS and websites and IoT information. On the other hand, SAP HANA is used for processing data from the operations system. By linking both systems, a wider range of analytics can be performed.
According to Ryohei Kubota, project manager of the Information Excellence Group, Information Systems Department, “While there were many factors that influenced our decision to select Hadoop, eventually it all came down to cost. We had instructions from our CIO to select a system that enables as much storage as possible, and we thought an open source Hadoop platform is most appropriate way for building a platform that can process large volumes of data efficiently at a lower cost.”
Other key factors that tilted the balance in Hadoop’s favour were its previous successful deployment in Japan and elsewhere, and Hortonworks’ offering of round the clock support. “When we first started considering the deployment with our consulting partner VUPICO, there were already several HDP deployment cases among leading beverage manufacturers in Japan, and this was a huge advantage. Those case studies were very helpful,” adds Shobu. “We help analyse data in order to increase business value for our customer. This platform requires a cost-effective and flexible technology approach that can support changing analytical needs.
"LIXIL decided to deploy HDP due to its globally proven results. In collaboration with Hortonworks Inc., we developed LIXIL’s information analysis platform,” says Vukojevic from VUPICO.
“Hortonworks is able to provide prompt support even for new technologies developed by the Apache Hadoop community. For example, the ORC file format is optimised for Hive processing. We were impressed by their high technical skills and constant support,” adds Kubota.
A wide array of different components are currently deployed at LIXIL. These include, Hadoop distributed file system (HDFS), YARN resource management capabilities, Hive SQL processing, Spark in-memory processing, and operation management tool, Apache Ambari. In addition, the deployment of an information analysis verification system is also underway.
Koshi Funamizu of Data Analysis G, Information Excellence Group, Information Systems Department echoed the positive sentiments regarding the use of HDP. “We are currently testing a variety of components as we have not decided their specific usage. However, we find Ambari very convenient. For example, it allows us to update configuration files on several servers simultaneously, and we can easily check YARN resources and HDFS disk space on GUI."