House Foods Corp., which constructed a data warehouse system for sales information,recently introduced an 8-way IA server to improve response. It also introduced a Symmetrix storage system, reducing response times to 1/3 compared to the previous system. Improved response has brought about visible effects in the form of a 100% increase in the number of monthly user accesses.

----The existing analysis system was no longer able to respond to changes in management conditions

Toshimoto Koshio Team Manager Information Systems Division, Planning Section

"House Vermont Curry"---if you've lived in Japan, chances are you've tasted it at least once. It was released in 1963, and immediately became a major hit, eventually making House Foods the top manufacturer of curry products in Japan. Based on the corporate concepts of "more delicious, more convenient, healthier foods," House Foods has expanded its range of product categories to include processed foods such as stews and packaged noodles, retort pouch foods, and even beverages and snack foods. The food industry environment is changing dramatically, however, with the extended consumer slump and intensifying competition in a market that has reached the point of saturation. Other factors include the recent problems related to BSE or "mad cow disease," and the resulting issues regarding product labeling.

In the midst of this complex environment, House Foods is conducting "Proposal-style sales," in which scientific POS and "planogram" analyses are carried out based on the latest data, and salesmen actively propose sales strategies to customers to ensure optimum management not only of House products, but of all products in the relevant categories. The information analysis system designed to support this sales approach is a TQF tool, introduced in 1994.

The TQF tool enabled users to search for data in a Relational Database Management System (RDBMS) on a mainframe using a wide range of conditions---and also to process this information---all from an emulator terminal. Because users were able to search for data themselves, the system was extremely effective in terms of sales support.
The only searchable data was that which had already been aggregated---that is, the user could not search for raw data---which meant that detailed analyses were impossible, and because the database was constructed on a mainframe, data analysis based on long-term trends was also unfeasible. "There were some analysis frameworks that had been viable when the system was first implemented," says Toshimoto Koshio, a Team Manager in the Information Systems Division Planning Section, "but with so many changes in the market and management environments, these frameworks were no longer able to accommodate the kinds of analyses we needed to perform. Even in terms of developing analysis items that would meet user needs, mainframe programming required to many processes and too much time---the system had obviously reached its limits." The company thus began considering the construction of a data warehouse from around 1998.

----Problems with response in the data warehouse system

Taro Shiraishi Information Systems Division, Planning Section

In preparation for the start of the first stage of development, House Foods received proposals from several vendor companies. After studying the proposals, it selected Red Brick Warehouse as the database engine, and Brio Enterprise Series as the front-end tool. According to Taro Shiraishi of the Information Systems Division Planning Section, there were a number of front-end tools proposed for the data warehouse system by the various vendors, but all of the database engines were IBM's Red Brick Warehouse. "We conducted comparisons and evaluations against the most popular database products, but Red Brick was the most impressive, because it was designed especially for data warehouse applications, and because of its high-speed loading functions."

Brio Enterprise, which was implemented as a database access/analysis tool, adopted two key components: Brio.Insight, which enables the user to create reports and analyses---including OLAP (On-line Analytical Processing), Pivot, Drill, and other analyses---from a Web browser; and OnDemand Server, which executes database processing as an application server via the Web server, and returns the results to the user.

In the first release (November 2000), the company constructed a database server using an IA server with a 4-CPU configuration and WindowsNT Server4.0 Enterprise Edition. This version was not able to achieve the required level of performance, however, in part because it was using an internal disk. Koshio describes the situation as follows: "The system wasn't able to complete batch processing, in which data was downloaded from the mainframe during the night, in time to provide the requested service, so response time was delayed even further."
In order to resolve the issue of response time, House Foods revamped the system in November 2001, with a focus on improving server performance and increasing disk I/O speed.

----Response dramatically improved with Express5800/180Rb-7 and Symmetrix

Teruhiko Kawasaki Information Systems Division, Planning Section

First, to resolve the problem of server performance, the existing hardware was replaced with the Express5800/180Rb-7, which allows an 8-CPU configuration, and the existing OS was replaced with the Windows 2000 Advanced Server. Then, a Symmetrix 3630 with a Fibre Channel was introduced to achieve faster disk I/O speeds. In addition to adopting a gigabit-level Ethernet for the network, response delays resulting from network problems were overcome by using Windows 2000 terminal services to access the WAN from sales offices with very limited bandwidths. As a result of this system renewal, response time was reduced to about 1/3 compared to the first stage release system.

Koshio made the following comments with regard to the replacement process. "We adopted the 8-way Express5800/180Rb-7 for the IA server because it offered the fastest performance in the Express series at that time. We also considered the Windows 2000 Datacenter Server for the operating system, but we got all the performance we needed when we switched from Windows NT to Windows 2000, so if we were only concerned about extensions to the data warehouse system, we decided there was no reason to go to the Datacenter Server." Nearly all of House Foods' hardware is made up of NEC products, from the ACOS mainframe to the Express5800 servers, clients, and PCs in each division. As for the reasons for choosing NEC products, Koshio says: "I'm sure that many users have been successful with multi-vendor systems, but it takes time to respond when a failure occurs---because it takes longer to isolate the problem, because it's hard to determine which vendor is responsible for repairs, and because you have to deal with so many different service contacts. With a single-vendor system, we can look forward to scale merits, and also to receiving NEC's comprehensive support."

----With improved response, user accesses increase by over 100%

The results of dramatically improved response and improved front-end tool operability are clearly visible in the form of increased user accesses. The sales information data warehouse is accessed from around 600 client PCs, which are located at the company headquarters and at nine branch offices and 36 sales offices across Japan. Before the renewal, the system was accessed about 4,500 times each month, but this number immediately jumped to between 8,000 and 10,000 accesses per month starting with the month following the system renewal. Even Koshio was surprised by this development. "It happened because we were able to respond to almost all of the users' requests for improvements, but we didn't expect that the number of user accesses would increase as much as they have."

By introducing the Symmetrix 3630, the company was able to establish a backup and development system that was independent of the main operation system, enabling development of the data warehouse using actual data, and data backups that didn't adversely affect services. TimeFinder, which runs on Symmetrix, allows data to be copied into the No. 3 volume, so data can be backed up quickly using this volume, and development can be conducted using real data even while services are being provided. "With the previous system, if we wanted to conduct tests using real data, we had to do it on holidays, explains Teruhiko Kawasaki of the Information Systems Division Planning Section. "With TimeFinder, we can develop the warehouse and test response using real data, and this has led to improved development speed."

Based on the corporate concepts of "more delicious, more convenient, healthier foods," House Foods produces and sells a wide range of products, including the major hit product "Vermont Curry" and other curry products and spices (condiments), processed foods such as stews and packaged noodles, prepared foods (retort products), beverages, and snack foods. Condiments and spices account for over 40% of the company's sales; House Foods has been the leading manufacturer of curry products in Japan for many years. The food product industry as a whole has been facing adverse conditions with issues related to BSE and the Food Sanitation Law, but House Foods has maintained integrity and customer trust by developing new products to meet a wide range of customer needs---such as for greater convenience, individual servings, and gourmet foods---and by ensuring even more meticulous quality management.