Nissan Cracks Down on Server Sprawl

July 14, 2010
Automaker cuts its use of servers from 250 to 44, lightening the load on its environmental and budget footprint.

Think of all the thousands of devices transmitting data nowadays out of a manufacturing facility -- from barcode readers, to RFID tags, to quality systems, to programmable logic controls (PLCs) -- shooting out millions of signals of information every hour. That massive increase in information has coincided with IT departments taking on exponentially more servers.

That, unto itself, has become a problem in many cases. Today, server sprawl -- that is, the vast expansion of physical servers that suck up large amounts of space and energy -- has grown at an alarming pace. Some have addressed this through coordinated efforts of virtualization.

For instance, Nissan North America Inc., watched as its servers sprawled to 250 units at its three vehicle manufacturing plants in Tennessee, in Smyrna, Decherd and Canton. But over the last two years, the company implemented a virtualization strategy for its assembly and component manufacturing division to bring the total down to 44.

That level of consolidation is notable, not just for its sheer scale, but also because the effort was carried out by the facilities' manufacturing and quality-assurance specialists -- not the standard company IT department, said Phil D'Antonio, manager of conveyors and controls engineering at Nissan's Smyrna plant.

The virtualization projects required bringing together dozens of pieces -- including a mixture of database servers, application servers and specialized servers, all of which send signals to more than 5,000 devices on the floor.

Over the past five years, Nissan's internal customers, such as production, maintenance and other departments, installed an array of quality, tracking and cost management applications, according to D'Antonio. Most of these applications couldn't exist on the same server. Though many servers had idle resources, new hardware had to be purchased at an alarming rate.

Nissan North America Inc. has
cut its energy costs by 34% by reducing the number of servers
at its three plants using
Microsoft Hyper-V virtualization
processing software."That's when we hit our critical mass," says D'Antonio.

Stretched out over three rooms, Nissan's server farm had swelled to 250 servers when the automaker decided to address the situation.

Following a thorough inventory, Nissan began by combining what might have previously been singular standalone manufacturing projects into one application, which was then run in one Microsoft Hyper-V virtualization processing system. Nissan managed to tackle two key goals: combining not only manufacturing applications, but also condensing its IT operations virtually.

"The problem we had was our servers were not used to capacity in many cases," says Matt Slipher, a systems engineer at Nissan. "And we couldn't really combine the servers because they had applications on them that didn't play well together. What Hyper-V does is it allows you to take one physical server and create many virtual servers, each with its own operating system.

"The underlying hardware is the same. It just allows you to use that entire server, where before we might have only been using it to 50% capacity." The consolidation has led to a 34% savings in computer electricity consumption at the three plants.

Equally important, the three computer rooms at each location have now been reduced to two. Separate computer rooms have to be maintained, regardless of virtualization efforts, so in the event of a mishap, one computer room wouldn't shut down the plant. Each room can serve as a recovery facility for the servers at another plant. Virtualization makes this process easier to maintain.

The Hyper-V systems are running manufacturing, quality-control, and process-engineering applications, such as paint mixing for different models. One added bonus is that fewer servers to maintain mean more man-hours to spend elsewhere, says D'Antonio.

"When you've got that many servers, you can imagine the amount of time it takes just to keep them running," he says. "Now, you can take those resources and focus them on value-add initiatives, like new applications and improving our cost and quality."

Of course, virtualization is not necessarily an answer unto itself. The ease with which administrators can provision virtual servers could belie the fact that the resources needed to support virtual machines can get expensive.

Applications, whether they are real or exist only in a virtual world, require processing, memory, storage and networking. In other words, virtual machines require resources and money.

Jett Thompson, a computing infrastructure architect at Boeing, spoke about the issue at a recent Gartner annual data center conference.

"If you don't have demand management and good governance in place, you're actually going to cost your company money," he warned. "Virtual server sprawl can wipe out any savings."

See also: Measuring What Drives Value Holistically

Sponsored Recommendations

Voice your opinion!

To join the conversation, and become an exclusive member of IndustryWeek, create an account today!