Faced with a massive PC refresh at a price tag of $1.8 million, Jack Wilson instead rolled the dice on virtual desktops three years ago. The enterprise architect at Amerisure Insurance didn't just dabble in the nascent technology, he enacted a sweeping change, replacing all 800 PCs with Wyse thin clients and a server infrastructure that hosts 800 Windows workspaces -- a feat that took eight months and, critically, struck at the heart of worker productivity in a services-dependent industry.

Wilson had to rethink his entire infrastructure, from servers and storage to the network and its points of potential failure. He had to convince employees to give up their powerful PCs and embrace thin clients. And he had to deliver a solid return on investment to his bosses. Essentially, Wilson gambled that desktop virtualization wasn't just a fad but the future.

[ Read InfoWorld Test Center's review of Citrix XenDesktop 2.0, which leverages streaming applications, server virtualization, and swift admin tools for a scalable and manageable virtual desktop infrastructure. ]?

Today, Wilson basks in the limelight of success. "I've been in IT for 30-plus years and had a lot of really great ideas, but none of them actually worked out as well as I had thought," Wilson says. Desktop virtualization "is the only thing I've done that has exceeded my expectations."

Bold companies such as Amerisure have embarked on the desktop virtualization journey, navigating uncharted waters and steamrolling toward a fuzzy future. Yet most are reaping the cool benefits of the technology today. Amerisure, for instance, expects to save a few million by avoiding next year's traditional PC refresh cycle as well as one in 2012.

The virtualization buzz The term "desktop virtualization" generally refers to centralizing desktop applications inside a datacenter. These applications can be hosted on a traditional server farm and delivered to remote users via Citrix or Windows terminal services, or in the case of VDI (virtual desktop infrastructure), they can be run in Windows virtual machines on a VMware server. End-users simply launch a browser on a thin client -- or any PC, for that matter -- and access their desktops over the network. Regardless of how you centralize the desktops, the essential benefits remain the same: Virtual desktops are easier to manage, provision, and secure than PCs.

On the dollar-savings front, Forrester Research estimates that desktop virtualization costs around $860 per user (plus any required network upgrades) to deploy, which is less than the cost of a PC. Amerisure reaped not only these PC savings but additional benefits as well. Help-desk calls, for instance, have decreased by 80 percent because thin clients rarely run into trouble. The cost of rolling out a new desktop operating system, such as Vista, is cheap with virtual desktops but a major hit on the balance sheet with PCs.

Another bonus: Wilson no longer frets about virus-laden PCs infecting his network or hackers stealing data from PCs. "I'm not worried about catching something off a laptop because I'm just borrowing the laptop's screen and keyboard," he explains. "In a traditional VPN model, it would be a problem because you would be connected to my network."

Companies are drawn to desktop virtualization for a variety reasons. Lifetime Products, a manufacturer of polyurethane tables, sheds, and basketball hoops, chose desktop virtualization for its inherent data protection aspects. Desktop virtualization ensures that all of Lifetime's product-design data and other intellectual property remain safely locked inside its datacenter in Utah, even though engineers scattered here and abroad work daily on new designs.

Last year, Natixis Capital Markets virtualized workstations for 30 employees (out of 400) to give them server-level reliability and horsepower. Natixis deploys virtual instances of desktops on servers inside its datacenter for its army of traders. "If some traders ran apps locally on their workstations, then they wouldn't be able to do anything else on them," says Drew Hiltz, CTO of Natixis in the United States.

Slow from the starting gate Despite all the buzz around desktop virtualization, there are signs of tepid adoption. In a recent survey conducted by sister publication CIO, only one out of four respondents was using desktop virtualization; one in five said it would be a year to three years before they'd deploy the technology; and 37 percent said they weren't interested at all.

Why all the hand-wringing? The fact is, desktop virtualization has a few technical blind spots that it still needs to cover. Graphics and streaming video don't work well on a virtual desktop without significant (and costly) network upgrades. "If you have high-graphic apps, this is going to be a kludgy environment to work in," Wilson says.

Certain applications also don't run smoothly on a virtual desktop, while some software licensing even forbids their use. "There are vendors selling applications who still want to resist," says Hiltz. "Vendors play with licensing models to squeeze more dollars out of you. I can't run Bloomberg on a virtual desktop based on the language of the licensing, even though technically I could."

Both Hiltz and Wilson worry that virtual desktop users will drain datacenter resources. Part of the problem is that users feel resources are unlimited in the virtual desktop model. Another issue is that management tools are not yet up to par, in terms of controlling CPU and memory usage for every employee. "I'd like to be able to throttle a user down," says Wilson. "While vendors say they have this ability, that's not really true yet."

Desktop virtualization adopters say these concerns aren't even the toughest hurdles when deploying the technology. They point to the need for massive infrastructure upgrades that wreak havoc on ROI, resistance from managers who are wary of hitching workers' productivity to a live Internet connection, and pushback from end-users who don't want to lose control of their workspaces.

Sweeping inftrastructure upgrades Wilson's infrastructure upgrades included a retooling of his network's points of failure. He used to have two T1 lines at every location, yet twice a year, AT&T would go dark and take out both of them. This wouldn't work in a virtual desktop environment, where "our base assumption is that you will always be connected," Wilson says. "Generally speaking, if you aren't connected, there's very little you can do anyway. Maybe you can write a Word document or work on an Excel spreadsheet, but even then you'll likely need to connect to a file server."

And so Wilson replaced a dedicated T1 line with cable or DSL (depending on the location) and added a Cisco router. Now he says he has three levels of redundancy on different technologies. Also, Wilson added 10 quad-core Dell servers running Citrix XenApp (formerly Presentation Server).

Lifetime installed a powerful MPLS (multiprotocol label switching) mesh network to accommodate the fast transfer of images its engineers needed to work on over thin clients. "Since virtual desktops are mostly used for WAN applications, make sure that your bandwidth and, more importantly, your latency are as low as possible," advises John Bowden, CIO at Lifetime.

Fast storage is another important factor for reducing latency. Fortunately for Natixis, the company had just bought high-end storage from EMC for its SAN. "Reliable, fast storage is key," says Florent Soland, manager of Windows services and virtualization at Natixis. "If you don't have it, it's going to take a lot more time to provision a workstation, and the end-user performance will be very bad."

Infrastructure upgrades throw a monkey wrench into the ROI that desktop virtualization was sold on. Add up the expense of new SAN storage, servers, virtualization software, a connection broker, and thin clients, among other costs, says Soland, "and it's more expensive to roll out 100 workstations in a virtual desktop infrastructure than 100 desktop workstations today -- we're still not there yet."

Au contraire, says Wilson. The reason ROI suffers, he says, is that most companies transition only a part of their workforce to virtual desktops. Indeed, Natixis plans to move 80 percent of its workforce to virtual desktops over the next couple of years. Lifetime's target is only 20 percent. This means they must support two computing models, contends Wilson.

"If you do this piecemeal, then you're only adding another layer and making the stack deeper and more complex," Wilson says. "The key is to deploy this strategically across the board." (To be fair, Natixis cites desktop virtualization concerns, such as licensing and apps that don't work well on virtual desktops, as barriers to a wholesale transition.)

My workspace, your server Everyone agrees, though, that getting end-users to embrace virtual desktops is the greatest hurdle fronting the technology -- that is, removing the "personal" out of the PC in the employee mindset. In fact, when Soland first brought the concept of desktop virtualization to his boss, Hiltz's initial fear was that users would balk. After all, angry employees and productivity loss are death knells for any technology.

Soland solved this problem by making Hiltz a beta user. Soon Hiltz realized that a virtual desktop looked and acted the same as his old PC, except that he couldn't put a picture of his children on the screen's background. Natixis traders also liked the virtual desktops because they suddenly had more processing power. Moreover, Natixis didn't replace their powerful workstations with thin clients -- traders could still work offline using applications running locally. "We didn't use thin clients," says Hiltz. "A thin client without a network connection is a boat anchor."

It's important that end-users feel like they are getting something out of the deal, says Wilson. That's why he gave employees 19-inch monitors, the largest screens available at the time. And he touted desktop virtualization's biggest end-user benefit: the ability to work anywhere. Now Amerisure employees can work from home a couple of days a week.

Still, many people view laptops as a tether to their personal lives -- not just for work -- and shun the locked-down, thin-client world order. But Wilson sees desktop virtualization as an enabler for them. "I can see a day within the next three to five years when I will hire someone, and he'll say, 'I don't want you to supply me with anything; I just want to connect my MacBook to your environment,'" Wilson says. "Our environment will totally do that, and it'll lower my costs even further because I won't have to provide him with a thin client."

Road bump ahead? Infrastructure upgrades and cultural acceptance are two game-killing guns leveled at desktop virtualization. But there are other technical wrinkles that still need to be ironed out, says Doug Dineley, InfoWorld Test Center executive editor. There's no question that the environment serves up a host of benefits, such as better security, he says, but the jury is out when it comes to potentially hampering worker productivity.

Dineley feels that end-users will need a way to work offline on a virtual desktop -- a way to check out a virtual machine and work on files without a network connection, a la Google Gears. "As we get further along, I suspect we'll start hearing some horror stories," Dineley says, "like unintentional denials of service that bring the blood of users to a boil."

Microsoft has seen the future of the datacenter, and oddly enough it's missing a roof.

The company's future datacenter design, which will be its de facto standard in five years, is a cross between an electrical switching station, an RV-park and the closing "warehouse" scene from the 1981 film Raiders of the Lost Ark .

[ Stay ahead of advances in technology with InfoWorld's Ahead of the Curve blog and newsletter. ]

The company envisions a set of prefabricated containers the size of a semi-trailer filled with as many as 2,000 preconfigured servers. The containers can be parked next to and plugged into pre-built mechanical, electrical, cooling and security components. In essence, it is a giant collection of boxes and pluggable components that can grow and shrink based on need.

The container portion of the idea is nothing revolutionary. Microsoft is installing them in its Chicago datacenter. Sun introduced a server container called Project Blackbox in 2006 and Google received a patent in 2007 on its "mobile datacenter" stored in a standard shipping container, which unlike Sun's Blackbox, could be clustered in the same modular fashion that Microsoft is proposing.

The container idea also has its critics who say they are rife with electrical and mechanical concerns, have power management and cooling issues, present a single point of failure, and are susceptible to damage during shipping.

Microsoft, however, is not just talking about containers, but the configuration of the entire datacenter.

The company this week unveiled what it is calling its "Generation 4 Modular Data Center" plan, a blueprint that will define its cloud datacenter infrastructure in the next five years.

The datacenters have four walls and a sophisticated perimeter security system, but are open to the elements as they lack a roof. Trucks wheel the boxes into the enclosure where they are connected to power/cooling stations before being brought online.

It's a bold plan to drive industry thinking about how to construct and operate datacenters in a world of capacity spikes, real-time needs for computing power and expanding green initiatives.

"We believe it is one of the most revolutionary changes to happen to datacenters in the last 30 years," said Michael Manos, general manager of global foundation services for Microsoft, in his blog introducing Microsoft's Generation 4 plan.

BusinessWeek reported last month that Microsoft said it was going to "reinvent the infrastructure of our industry" by building some 20 datacenters that can carry a price tag as much as $1 billion apiece.

"In short, we are striving to bring Henry Ford's Model T factory to the datacenter," Manos said. "In our design process, we questioned everything. You may notice there is no roof and some might be uncomfortable with this. We explored the need of one and throughout our research we got some surprising (positive) results that showed one wasn't needed."

Microsoft says all the pieces needed to construct the datacenter would be built off-site and assembled once they arrived at the datacenter location, much the way planes, cars, and computers are built today. The company says the process would mean less time and money to erect a new datacenter.  

And Microsoft expects efficiencies in power usage that blow away even the best-rated facilities today based on PUE (power usage effectiveness), a metric developed by The Green Grid and used to determine the energy efficiency of a datacenter.

"A key driver is our goal to achieve an average PUE at or below 1.125 by 2012 across our datacenters," Manos said on his blog.

Achieving such a low PUE average would be a breakthrough given that the typical datacenter has an average PUE of 2.5, according to the Uptime Institute. The Institute says that a best-case scenario today could produce a 1.6 PUE average if the datacenter is using the most efficient equipment and best practices.

Manos added: "More than that, we are on a mission to reduce the overall amount of copper and water used in these facilities."

Operationally, the datacenter would offer different classes of service defined to meet the needs of applications and services deployed and to create cost efficiencies. The classes include ala carte options, such as uninterruptible power supplies and backup generators, temperature controls, and redundancy.

Microsoft says the varying configurations will drive engineering innovations that will lower operational costs for applications.

And to show that Microsoft is aiming toward industry interoperability the datacenter's containers would have common interfaces so others can plug their wares into them including computer vendors, UPS vendors, and generator vendors.

Manos admits that a 2005 memo written to Microsoft employees by now chief software architect Ray Ozzie was the trigger for thinking about how Microsoft would evolve deeper into an operations company rather than a provider of packaged software.

What grew out of that were Generation 2 facilities, now operating in Quincy, Wash., and San Antonio, Texas, which took into account sustainability, energy efficiency, and total cost of operations.

Generation 3 facilities, which are represented by Microsoft's mammoth datacenter in Chicago, feature containers and a modular design.

Microsoft has posted a short video to show how its Generation 4 datacenters would be constructed and how they would operate.

Network World is an InfoWorld affiliate

With credit in a tight squeeze and the economy in free fall, the next few years should see the collapse of some small PC makers and a restructuring of the rest of the industry, according to an industry research firm.

Richard Shim , personal computing research manager at IDC, told Computerworld that he expects a consolidation of the market. However, he doesn't think that the big PC players, like Hewlett-Packard , Dell, and Acer , will gobble up smaller hardware vendors. Instead, he said those smaller players will simply fold up shop as the faltering economy keeps companies and individuals from buying new computers.

[ Special report: IT and the financial crisis. ]

"It won't be so much about acquisition, but the smaller players will just go away," said Shim, adding that he thinks the industry could lose fewer than 10 companies. "The big players are feeling the hurt as well. Right now, everybody is beating each other up in price. If some are going to die off anyway, what's the sense in buying them?"

Just yesterday, IDC reported it is projecting that worldwide PC sales will quickly drop off because the sagging economy is causing people to hold onto their savings while credit is unavailable. IDC noted that it expects PC shipments to inch upward by 3.8 percent in all of 2009, but added that the values of those shipments will drop by 5.3 percent. In the U.S., expectations are bleaker as IDC predicts that shipments here will decline by almost 3 percent in 2009 with low single-digit increases in the next few years.

Shim noted that he thinks MPC kicked off the consolidation early in November when the PC maker filed for bankruptcy . In mid 2001, the business was sold by parent company Micron Technology, and later changed its name from Micron Electronics to MPC. A little more than a year ago, MPC bought Gateway's professional business division.

And MPC won't be the only mid- to low-tier PC maker to fall, according to Shim. He said 2009, 2010 and maybe even 2011 will be tough years for the industry.

Just a few weeks ago, iSuppli slashed its 2009 growth forecast for worldwide PC shipments by nearly two-thirds because of the deteriorating economy. The analyst firm is now projecting that worldwide PC shipments will rise by 4.3 percent in 2009, down from its previous forecast of 11.9 percent growth. iSuppli also adjusted its expectations for 2010, dropping its initial prediction of 9.4 percent growth to 7.1 percent.

ISuppli's adjustment to its PC forecast came on the heels of the firm downgrading its estimates for global semiconductor revenue for the year. The researcher projected that 2008 semiconductor sales will decline by 2 percent to $266.6 billion, from about $272 billion in 2007. In October, ISuppli had predicted that 2008 semiconductor sales would grow by 3.5 percent over last year's. The researcher also predicted that the negative momentum will continue into the fourth quarter of this year, with the overall market expected to drop by 10.9 percent compared with the same quarter last year.

Shim noted in an interview on Thursday that the economic crisis likely will reset many vendors' business models.

"If you change selling prices, you need to find new ways to do business, like skimping on quality or coming out with new models," he explained. "A lot depends on how drastic the situation gets."

Computerworld is an InfoWorld affiliate.

Google called off its proposed search advertising deal with Yahoo just three hours before the U.S. Department of Justice was to file an antitrust complaint on Nov. 5 aimed at blocking it, according to the lawyer that the government hired to pursue the case.

In an interview with the legal blog AMLaw Daily published Dec. 2, Sanford Litvack -- the attorney who would have been the lead counsel on the antitrust case -- said that Google and Yahoo decided to abandon the proposed deal shortly after DOJ officials informed them of the agency's plans to file the antitrust complaint.

[ In the aftermath of the failed Google-Yahoo deal, Jerry Yang resigns, and the Microsoft-Yahoo dance continues. | Keep up on the latest tech news headlines at InfoWorld News, or subscribe to the Today's Headlines newsletter. ]

Shortly after the deal -- which would have had Yahoo running Google advertisements alongside its own search results -- was announced in June, Google and Yahoo came under fire from large advertiser groups, which charged that the arrangement would diminish competition and raise online advertising prices.

And an antitrust think tank said the partnership could end up as a "black hole that swallows up Yahoo," thus justifying an antitrust investigation.

In addition, the chairman of the U.S. Senate's antitrust subcommittee in October urged the DOJ to closely examine the proposed partnership, noting that it could lead to higher advertising prices and create unfair market conditions. Sen. Herb Kohl (D-Wis.) said at the time that the subcommittee's investigation found that many advertisers and competitors were concerned that Google would control a dominant share of the search advertising market. Under the deal, Yahoo would have less incentive to compete against Google and could opt to exit the market altogether, Kohl asserted.

Litvack went on to note that the complaint -- which was never filed -- would have charged that the deal would have violated antitrust regulations that bans agreements that restrain trade and prohibit companies from monopolizing or attempting to monopolize trade.

"It would have ended up also alleging that Google had a monopoly and that [the advertising pact] would have furthered their monopoly," Litvack said in the blog. "The fact that we filed a lawsuit would not by itself have stopped them. We would have had to get an injunction from the court, and we would have sought that."

In the interview, Litvack went on to acknowledge that Microsoft Corp. and other companies lobbied the department to block the proposed deal, but said that the efforts had no influence on his decision to recommend that DOJ block the deal.

"[The department is] making it clear to the parties and to the world that this is how the division viewed these particular aspects of Google's business," Litvack added in the blog. In a statement after the deal was called off the DOJ said that under the agreement, Google and Yahoo would have become collaborators rather than competitors for a significant part of their search advertising business," materially reducing important competitive rivalry between the two companies."

Computerworld is an InfoWorld affiliate.

A new computing fabric to replace today's blade servers and a "pod" approach to building datacenters are two of the most disruptive technologies that will affect the enterprise datacenter in the next few years, Gartner said at its annual datacenter conference on Wednesday.

Datacenters increasingly will be built in separate zones or pods, rather than as one monolithic structure, Gartner analyst Carl Claunch said in a presentation about the Top 10 disruptive technologies affecting the datacenter.

[ For recent news on the modular design approach to datacenters read "Microsoft applies Model T factory methods to datacenters." ]

Those zones or pods will be built in a fashion similar to the modular datacenters sold in large shipping containers equipped with their own cooling systems. But datacenter pods don't have to be built within actual containers. The distinguishing features are that zones are built with different densities, reducing initial costs, and each pod or zone is self-contained with its own power feeds and cooling, Claunch says.

Cooling costs are minimized because chillers are closer to heat sources; and there is additional flexibility because a pod can be upgraded or repaired without necessitating downtime in other zones, Claunch said. (Read more about how to reduce cooling costs in the datacenter.)

"Modularization is a good thing. It gives you the ability to refresh continuously and have higher uptime," Claunch said.

By not treating a datacenter as a homogenous whole, it is easier to separate equipment into high, medium, and low heat densities, and devote expensive cooling only to the areas that really need it, Claunch added.

The move to pods and zones is among what Gartner calls the most disruptive technologies affecting the datacenter. In no particular order, these technologies are storage virtualization; cloud computing; new server architectures; PC virtualization; enterprise mashups; specialized systems (aka hardware appliances); social software, and social networking; unified communications; zones and pods; and green IT. 

Many of these technologies have been covered by Gartner in previous lists (including "Gartner's Top 10 strategic technologies for 2008" and "10 strategic technologies for 2009"). Enterprises won't have to wait long to take advantage of these technologies: All these trends are beginning to happen now or will do so within the next few years, Claunch said.

If Gartner's predictions are correct, the server industry is soon to undergo a significant transformation.

Gartner views today's blade servers as an interim technology that will give way to a new, more flexible type of server that treats memory, processors and I/O cards as shared resources that can be arranged and rearranged to suit a business's needs. Like virtualization technology, this computing fabric of the future will make hardware more adaptable to changing needs.

IT shops will be able to create machines of whatever size they need, and shift resources around as often as necessary, Claunch said. In addition, instead of relying on vendors to decide what proportion of memory, processing and I/O connections are on each blade, enterprises will be able to buy whatever resources they need in any amount, a far more efficient approach.

While rack servers are self-contained units, today's blade approach allows a combination of some components, Gartner notes. (Compare blade servers.) I/O cards don't have to be included in each blade because they are accessed over a shared fabric. Memory and processors are still fixed parts of each blade, however, limiting flexibility. If extra memory is needed, you may have to buy another blade instead of just accessing the memory of another one, Claunch said.

"The next step in this progression is the introduction of technology to allow several blades to be merged operationally over the fabric, operating as a larger, single system image that is the sum of the components from those blades," a Gartner PowerPoint presentation states. "The fabric-based server of the future will treat memory, processors and I/O cards as components in a pool, combining and recombining them into particular arrangements to suit the owner's needs."

For example, an IT shop could combine 32 processors and any number of memory modules to create one large server that appears to an operating system as a single, fixed computing unit. This approach also will increase utilization rates by reducing the resources wasted because blade servers aren't configured optimally for the applications they serve. "This evolution will simplify the provisioning of capacity to meet growing needs," Gartner states.

Network World is an InfoWorld affiliate.

Advocates of net neutrality rules in the U.S. have called on President-elect Barack Obama to act quickly to prevent broadband providers from blocking or impairing access to Internet content of customers' choice.

The Open Internet Coalition on Wednesday called on Obama to follow through with his promises during the presidential campaign to establish net neutrality rules. Members of the coalition also called on Obama to appoint a new chairman of the U.S. Federal Communications Commission who would enforce net neutrality rules and champion broadband competition.

[ For an in-depth look at the battle lines that have been drawn over net neutrality, check out InfoWorld's special report. And for the latest in government IT news and issues: Subscribe to InfoWorld's Government IT newsletter . ]

In addition, Obama should appoint leaders at the U.S. Federal Trade Commission and Department of Justice who will promote an open Internet through antitrust and consumer-protection laws, and he should put key staff in place at the new office of U.S. chief technology officer and the Office of the U.S. Trade Representative to promote open Internet ideals both in the U.S. and overseas, the groups said in their letter to the Obama transition team.

"Net neutrality is certainly one key part of the platform, but it's only one part," said Art Brodsky, communications director for Public Knowledge, a digital rights advocacy group that's part of the coalition. "We want to make certain that these institutions of government are prepared to implement an open Internet agenda and respond appropriately to threats."

Members of the group, asked if they believe Obama will act on net neutrality and broadband competition given priorities such as the U.S. economy, said they expect the new president to move ahead on tech issues. "Providing affordable, accessible, high-speed Internet to all Americans is part of the economic recovery," said Markham Erickson, director of the coalition.

Coalition members said they're encouraged that Obama talked about keeping the Internet open in a tech policy paper released more than a year ago. "When it comes to people lobbying, we're always going to be behind," Brodsky said. "The telephone companies are the biggest, wealthiest, and have the [largest] army of people out there. What we have now that we didn't have before is backing of an administration."

Large broadband providers have questioned the need for new net neutrality laws, saying that the FCC has already acted against carriers that have unreasonably blocked or slowed Internet content. Strict net neutrality rules may discourage investment in broadband networks at a time when Obama is calling for more broadband, they have said.

A representative of Comcast, the nation's largest broadband provider, declined to comment on the coalition's letter. Representatives of Verizon Communications and Hands Off the Internet, a group opposing net neutrality rules, weren't immediately available for comment.