SERVING UP POWER

KIM ANN ZIMMERMANN

Most supermarkets have major investments in mainframe computers and powerful servers and rely on them to perform a host of crucial functions, such as financial and human-resources applications.

At the same time, supermarkets are striving to provide store managers with the data and transaction power to serve the customer efficiently and effectively.

Retailers and wholesalers are leveraging their mainframe, legacy and server investments in innovative ways in the era of networked personal computers and thin client servers.

It is a delicate balancing act to make sure store managers have the computing power to make on-the-spot decisions while the headquarters staff has the computing muscle to store and analyze two years' worth of sales data, for example.

"We're looking for hardware solutions that will provide us with the nimbleness we need to run our business effectively," said Mathew Francis, director of information technology for Save Mart Supermarkets, Modesto, Calif.

"We are making decisions about what the best software applications are, based on our business objectives. From there, we look at the hardware and networking issues," he said. "We've made a conscious decision to structure our systems so that we are not locked into one hardware platform."

To gain the desired flexibility, Save Mart, like many supermarkets, distributes its computing power throughout the organization rather than relying on one mainframe or super server as a repository for its data.

"Our strategy is to have several data repositories and use the right tools to pull that data together for decision support," he said.

A distributed computing environment can be more costly to maintain than a centralized mainframe or large server because adding hardware to the network increases the likelihood that something will go wrong. However, a distributed computing environment provides retailers with a greater array of software options and the ability to put computing power where it is needed.

"Yes, there are more boxes, but they are more robust," Francis noted. "We are willing to look at a minimally higher operating cost if we get higher performance and higher return on the investment we make," in terms of enhanced performance, features and functions, he noted.

"What we're seeing in many companies, including supermarkets, is that the mainframe is being surrounded by a number of servers, doing divisional and high-end departmental tasks," said Dan Kusnetzky, program director for operating environment and serverware for International Data Corp., a Framingham, Mass.-based research firm. "The multi-tier, multi-operating system strategy is key to survival."

John Granger, chief information officer for Furr's Supermarkets, Albuquerque, N.M., noted that there are many attractive alternatives to the mainframe that can provide flexibility for outsourcing support functions.

"With the mainframe, we pretty much had to stay in-house. As we're moving to the client-server environment, we can selectively outsource support for a number of applications. We don't need a mainframe specialist, and we can have someone working on just one part of our system," he said.

Furr's is in the process of moving applications off the mainframe to a client-server network. The chain has almost completed a project to move financial and human-resources functions off the mainframe and plans to have all its core business applications off the mainframe by the end of the year, Granger said.

When making hardware decisions, retailers are looking at several key factors, including who needs access to the data, how much data they need to access and how quickly they need to view that information.

At Pay Less Supermarkets, Anderson, Ind., store managers can access real-time sales data that is stored on store-based servers. That information is uploaded daily to an even more powerful mid-range computer at headquarters.

"The people at the stores need real-time access to data when making decisions," said Paul Nicholson, vice president of finance and management information for the chain. "If we're having a sale on strawberries, for example, the store manager needs real-time access to real-time sales data to make decisions about reordering. At headquarters, they don't necessarily need to know how many strawberries were sold at a particular store by 11 a.m. on a Tuesday. They need power to crunch the numbers for all the stores and don't necessarily need access to real-time item-movement data," he noted.

At headquarters, they need the hardware power to store historical data and analyze trends, Nicholson noted. "They need to analyze the impact of promotions, look at data from competing retailers, and to do that accurately you need to keep historical data, which is where our mid-range server comes in," he said.

Nicholson, echoing the comments of other retailers, noted the need to constantly evaluate hardware decisions as technology evolves.