For Patel of GMAC Commercial Mortgage, cramming applications onto servers wasn't something he could sell to management. "You can't get management excited about consolidating servers," he says.

Instead, Patel focused on setting up a network where he could slice and dice bandwidth just as other companies slice and dice processing power on servers.

For instance, Patel at the end of the month can allocate 10% of his network to voice traffic, 10% to video conferencing, 20% to data and the remainder to handling financial statements, when the rush comes for figuring results.

Giving priority to certain tasks is something management notices. "People notice when the connections are right," Patel explains. "Service feels faster, and remote locations can get set up faster. It's a win that has management asking for more."

There are two tangible benefits. With better control of bandwidth, GMAC has begun to service mortgages for other companies on a private-label basis. Another payoff comes with the flexibility to open new offices in Boca Raton, Fla., or other cities where demand for mortgages is surging.

By simply plugging those offices into GMAC's existing voice-over-Internet Protocol network, Patel can start calling services quickly while bypassing telecom service providers. The savings last year: $160,000, just on international long distance calls.

And GMAC never knows when it will need more capacity to handle all the PDFs, digital signatures and wire transfers around the country that come when it has to move $20 billion in mortgages in a day. So Patel can click on more bandwidth, when needed, through a Web site hosted by Masergy, a young network services provider based in Plano, Texas.

"We wanted to throttle up bandwidth ourselves, and we didn't want to call and then have to wait a month for someone to come out," Patel says.

In the future, Patel hopes the slicing and dicing of bandwidth can be automated as well. The network should "know" when the digital files associated with $20 billion worth of mortgages are about to hit—and adjust accordingly.

Tim Howes, Opsware's chief technology officer, says automating networks is arguably the most important piece of getting a virtual infrastructure in place.

But the tools for watching effects on other parts of a network or networks are not there yet. "Today, if an administrator takes down a router, he doesn't know what customer he's affecting and whether it's a financial or customer relationship management application that's impacted,'' Howes says.

That means operators of corporate networks have to figure out just what aspects of network and data center operations need most to be measured and then establish baselines for what costs and service levels should be.

"Ideally, you should know what it costs" to make sure an application only goes down for one minute a month, says Michael Mullany, vice president of marketing at VMware. That's the current standard, known as 99.99999% availability.

Then, figure out what it costs to take that to one minute every 10 months. "You could figure it out in a spreadsheet after a couple of days, but that information isn't readily available today and integrated with labor and asset costs,'' Mullany says.

Figure it out. And plan accordingly.