Linux Checks In

What seemed like a mundane purchase at the Hannaford Bros. supermarket in Standish, Maine, on February 4, 2003, was actually anything but ordinary. It was a milestone marking the first grocery transaction in the country using a Linux-based point-of-sale system.

To understand the significance of the milestone, it is important to place it in the context of frequently heard questions surrounding the Linux operating system: Is freely available, open-source software the right choice to run a billion-dollar POS supermarket chain? Is an OS created by a visionary student in Finland, Linus Torvalds, and supported by a voluntary group of freedom-fighting programmers the best platform to make a seven-figure capital investment? And, finally, although Linux is everybody's favorite penguin is it really ready for prime time?

To the latter question, the answer most IT executives give is a qualified "no." (See "What Others Are Saying about Linux" on page 17.) Most retail execs, in fact, believe Linux is a modest operating system with much to be modest about. While it will acquire many more applications and support resources in time, most believe Linux is still several years away from widespread adoption by major retailers.

Obviously, Bill Homa, Hannaford's senior vice president and CIO, respectfully disagrees. Homa's rollout will be installed in a third of the company's 120 stores by this fall and then, after a pause during the Christmas selling season, full deployment will be completed in the first half of 2004.

Why Linux POS? Why now? And why is Homa willing to take what many experts believe to be a huge gamble? Here's the inside story.

Time for a Change

Retailers dislike replacing POS systems. Doing a major overhaul once every 10 years is considered often enough. In late 2000, Hannaford's time arrived. The software vendor, ICL (now Fujitsu), informed Homa it would no longer support the retailer's aging ISS 400 POS after January 31, 2002. Like it or not, Hannaford needed to go shopping for a new POS.

"The existing point of sale worked fine," says Homa. "It was just being obsoleted. When a company announces it is dropping a product, developers have no incentive to stay around, so the cost of making changes goes up dramatically. Also, when something is going to be obsoleted, you basically stop making investments in it and this takes a toll over time."

So, about two and a half years ago, Homa and a small IT team consisting of Kevin Green, director of retail support, and Kevin Carleton, director of retail systems, began making plans to replace Hannaford's POS. Since it was going to be a major investment that would last up to 10 years, Homa wanted to leapfrog current technology and build an architecture to position Hannaford for the future.

"We looked at offerings by the larger vendors, IBM and NCR, and, honestly, we weren't thrilled," Homa recalls. "We started thinking that we had todo something very dramatic. We thought this might be the timeto start integrating point of sale into the rest of our architecture. And by thatI mean, making it highly centralized and Web-based, and moving servers out of the store."

Homa explains that most retail operations include five to six servers per store, and the reason is to compensate for slow and unreliable telecommunications. In this scenario, thin-client technology has some negative attributes because a centrally located server puts you at the mercy of the network.

But telecommunications have come a long way in the past few years. Going forward, Homa envisions Hannaford stores running on a single Microsoft Windows 2000 or Linux server that will store all the prelogs and host all the lanes. No security blanket in sight.

Doesn't it take a level of guts to trim down to one server per store? "Yeah," responds Homa, "But security blankets drive costs." In other words, multiple servers were the only way to run a chain store several years ago, but advances in the world of telecommunications have changed the paradigm.

"You tended to put lots of servers and lots of technology at the store level to compensate for limitations in the network," says Homa. "Nowadays, you have high-speed networks that are fast and reliable and inexpensive. You don't need a bunch of servers in every store, because that's an expensive way to go. You can pull technology back centrally and go with a thin presence in the store."

Homa was already centralizing technology in other parts of the store and decided to use the same approach with POS. Developing a POS criteria list was the next task, and it included finding a package that more than matched the functionality of the old Fujitsu software and thin-client POS with hardware that has no moving parts, meaning no fan or hard drive. Hannaford's wish list included embedding a version of Linux OS in the lane terminal and integrating it into the rest of the store's centralized architecture. Easier said than done.

With the criteria list in place, Homa began looking for vendors and quickly discovered "nobody really had what we wanted." The problem was that Homa was inventing something vendors had never seen before.

Finding Partners

After running into dead ends with several leading POS hardware vendors, plus several PC manufacturers, Homa found partners willing to satisfy the criteria — Wincor Nixdorf for hardware and Retalix for software.

Illustrating how this type of partnership can work for mutual benefit, Homa joined forces with Wincor Nixdorf to devise a POS terminal that has no moving parts. A fan in the power supply is the only moving part and it is built into an external brick that sits outside the system.

"The beauty of the system," explains Homa, "is that there is almost nothing to break. We still have one fan, but it's away from the POS terminal and is very easy to replace. We also have something that is totally USB configured, so all peripherals have a common interface. The theory behind this is that the cashier, if necessary, can swap a printer. If the whole thing is USB, it doesn't matter what port you plug into. With terminals like this it's hard for something to go wrong."

After being turned down by several hardware vendors "Wincor Nixdorf basically invented what we wanted," says Homa. "And since they had a ton of Linux experience, they provided a lot of guidance to Retalix, which had a great POS application but no Linux experience. So, it was a great collaborative effort between two vendors that supported and reinforced each other."

The benefit to the vendors, according to Homa, is that the products built for Hannaford represent a proof of concept that will serve notice to the rest of the industry and lead other retailers to embrace them.

Homa chose to go with Wincor in August 2002 and Retalix soon after. In the end, they were the only ones that could meet the criteria.

The Roll Out

A little less than a year after the primary partners were selected, a Hannaford customer in Standish, Maine, made a Linux POS transaction at 7:00 a.m. on Tuesday, February 4, 2003. The cashier had less than two hours of training on the new system, but everything went smoothly. But to reach that point, a massive effort took place behind the scenes to pull together last-minute details, culminating in a late-night installation session that ended about an hour before the store opened.

Aside from solving leading-edge hardware and software issues, Homa also had to contend with peripherals that had yet to develop Linux drivers. These included Epson for the printer, NCR for the scale/scanner, Symbol for the hand scanner and Hypercom for the credit-card reader. "They knew it was coming," says Homa, "but we pushed them to make the final development. In some cases, they didn't have a USB version. One vendor had to develop a USB version and a Linux driver at the same time. And that wasn't done until December, so we're about six weeks away from going to our new store and one of our major components is still sending us beta codes."

Three days before going live, the equipment arrived at the store and preparations were made for a Monday night installation. "This was such a big deal that at one point, around midnight on Monday night, there were 30 field technicians from eight different vendors in the store," Homa recalls. "Everyone that had a piece of the project was there to make sure their piece worked as planned."

Everything did work, and the initial pilot project was off and running. The second pilot installation is scheduled to take place the first week of May. By the end of June, the pilot phase will end and full rollout will begin in July.

Without a Net

Unlike most major IT projects, the Hannaford POS replacement didn't require a detailed business case or return-on-investment analysis to justify the expenditure. The old software was being obsoleted. Time to go shopping. Allocate the necessary capital. Move onto the next state.

But in choosing the Linux operating system, the 7.2 version offered by Red Hat, Homa added a high level of risk to a project that could have been an extremely safe bet. The decision couldn't have been easy, because Linux is new in retail and still intimidating to many technologists and senior-level executives. It isn't owned by anyone and retailers are forced to rely on the help of an ad hoc, open-source community.

Still, the hesitation many IT execs feel toward Linux is unjustified, believes Homa. "When you start to work with Linux," he says, "you understand that it is really a form of Unix. A lot of retailers have experience with Unix and they adopt it in about five minutes. The learning curve for my IT people was essentially nothing. If you understand that it is a first cousin to Unix you are over a big psychological hurdle."

Hannaford's IT staff was already somewhat familiar with Linux after installing scale-management system and Web-camera security systems that run on the open-source OS. They knew it was easy to use, scalable for the future and, as others had discovered, very stable.

In the end, one of the primary reasons why Hannaford decided to invest in Linux POS is contained in just four words: total cost of ownership. "The initial cost to us will be far less than anything else that's on the market," Homa says, "and then the overall total cost of ownership over the system's lifetime will be far less, too, because our new POS system is easy to maintain and Linux won't obsolete the hardware."

Many questions surround the adoption of Linux in retail, but as Homa discovered, the answers are out there if you look.