Over the past two years, Google has spent a whopping $5 billion to construct a network of state-of-the-art data centers. The company’s goal is to meet not only its own insatiable need for computing power, but yours as well.
Google’s investment is part of a broad movement toward “cloud computing,” a literally hazy term for a concept that takes outsourcing, software as a service, and similar rent-don’t-own trends to their logical conclusion. Think of it as a ubiquitous Wi-Fi hotspot that can satisfy all of a company’s computing needs, from software applications to data to communications and collaboration. No more data centers; everything you need is “out there” in the cloud, accessible for a fee. Or fees.
Your company may already be migrating to the cloud without your even knowing it. A recent McKinsey & Co. survey found that companies are increasingly moving away from being owners of IT platforms. Companies are spending 19 percent of their software budgets on applications delivered as services, the survey said. That is, they are eschewing the traditional model of licensing software for use on a company server in favor of subscribing to software that is housed on a vendor’s server. What’s more, 23 percent of executives think cloud computing is the best model for running IT systems outside a company’s four walls.
Cloud computing is a rebranded form of “utility” computing, a less-clever but more-apt description for the underlying model: computing becomes a utility that you simply plug in to. You would no more think of owning your own IT infrastructure than you would of owning your own power plant. Many of the products and services needed to achieve this vision exist today, but, as Frank Gillett, a principal analyst at Forrester Research, has noted, “The rising buzz about clouds is creating a cacophony of confusion.”
Cloud computing is not about handing off management of a data center. Instead, it’s about buying standardized IT capabilities — memory, storage, processing power, software — on-demand, over the Internet, billed according to usage. Companies could ditch some or all of their servers and other fixed assets and tap a virtualized pool of computing resources via an interface as simple as a Web browser.
“You can access a server with the exact capacity for a given application,” says Simon West, chief marketing officer of Terremark, an IT infrastructure provider. “When you don’t need the server anymore, you delete the configuration” from your agreement, thus seamlessly evading a common and costly IT pitfall, buying too much capacity.
Beyond its potential for greater economy, however, cloud computing has the potential to give business units faster, more-direct access to IT resources — one of business computing’s holy grails. “Cloud computing is really tackling how we manage the compute resources being utilized by a set of people,” says Dennis Quan, a director of development in IBM’s Tivoli division. “It represents a new level of efficiency in running and managing IT.”
A CFO may well ask how far this concept can go: Could cloud computing drive IT capital expenditures down to zero? Why buy a truckload of servers, after all, when the company can tap into Amazon or Google’s vast data centers and just buy processing power on an operating-expense basis? “The CFO is responsible for helping the company get the biggest bang for its buck,” says Jose Granado, a partner in the IT advisory group at Ernst & Young. “With cloud computing, companies — especially small and midsize companies — can realize significant cost savings.” The intense transactional workloads and other facets of large-company IT operations may not be as well suited to cloud computing — at least, not yet.
But even companies in the business of selling IT services are buying what they need via the cloud computing model. Consider Hoodiny, a Miami-based company that provides record labels with on-demand streaming capabilities for music downloads and other high-bandwidth services.
Hoodiny buys a bundle of processing and storage resources monthly, says Jose Gonzalez, executive vice president of business affairs, and can then configure virtual machines according to its needs. If the company needs to add five servers to handle a surge in additional Web traffic, it can bring new capacity online without a capital outlay or cannibalizing an existing server. And it can switch off the extra computing power when the promotion ends. “It takes 10 minutes to have a virtual machine available versus hours or days, depending on your financial position, to have the equivalent hardware ready,” Gonzalez says.
While Gonzalez is still collecting data from his five-person IT department to calculate the firm’s ROI, “instinctively you can begin to see there is some cost savings in cloud computing,” he says. The more immediate value, though, lies in eliminating the time it takes to decide to buy a new server, configure it, and go through purchasing, only to have the asset sit idle or underutilized after it fulfills a temporary need. “From an asset-management perspective, cloud computing is very attractive,” Gonzalez says. In the future, Hoodiny may rely on cloud services to replicate its production environment as a way to test upgrades and new software functionality. “I’ll have a duplicate platform without shelling out for additional sets of hardware” or having to beg for money, Gonzalez says.
Rebalancing Act
How useful will cloud computing be? For the most part, “companies are still struggling with it,” says Major Horton, CFO of Nirvanix Inc., provider of a cloud storage platform. “They say, ‘We get the advantage of not having to layout capital, but what do we use cloud computing for?’ That’s where there is still a learning curve.”
Heavy-duty enterprise jobs are not an option at this point. “The best [service-level agreement] that I know of provides for only 99.5 percent uptime, whereas most enterprises are looking for 99.999 percent uptime or better for mission-critical applications,” says Brett Waldman, a research analyst in system software at IDC.
Migration to the cloud will likely follow the same evolution other forms of outsourcing have taken: low-priority business tasks will be the early candidates. Instead of moving to cloud computing “en masse,” companies will “look at their technology landscape and rebalance it by pushing some to the cloud,” says Daz Wilkin, a program manager in Microsoft’s platform strategy group. The technical infrastructure that is not part of a company’s “secret sauce” will be best suited to cloud computing, Wilkin says, citing messaging and Web collaboration as examples.
Microsoft, for example, expects 50 percent of Exchange mailboxes to migrate to Microsoft Exchange Online in five years. “Customers are giving us jobs they cannot do themselves or that are peripheral and noncritical,” agrees Eric Novikoff, former CFO and now chief operating officer at ENKI, a virtual data-center provider.
There are good reasons for that: pivotal issues have to be addressed before companies are confident enough to shutter their own data centers in favor of Google’s or Amazon’s. It’s not clear, for example, that companies will always save money. Shifting the storage of unstructured data, such as multimedia libraries, to the cloud can reduce costs, Nirvanix’s Horton says. But a CFO would have to compare the total cost of owner shipper gigabyte of storage, throwing in expenses such as the percentage of building maintenance, physical security, and even property taxes allocated to storage systems, to make a valid comparison.
“It’s a highly individual calculation,” says IDC’s Waldman. “But what can be said is that cloud computing drastically reduces the start-up costs for new companies and projects, since there is no need to acquire hardware.”
Pricing and billing models will continue to evolve. Right now an apples-to-apples comparison is difficult. ENKI, for example, charges customers by hours of CPU time, gigabits per second transferred, and disk space used, while Terremark levies a monthly fee for a pool of computing resources. And even cloud computing’s largest ostensible appeal, a huge reduction in capex, can be a liability. Some customers are “resistant to switching their financial models from lump-sum expenditures to a flow-through-cash basis, often paid with credit cards or ACH,” Novikoff says.
In addition, for some applications, like accounts payable, a CFO’s priority may never be cost savings, points out James Zubok, CFO of Brainware, a developer of business-intelligence software. “The important thing,” he says, “is to have the information auditable and correct. CFOs are less concerned about how it gets done than that an invoice gets paid on time.”
The Not-So-Silver Lining
Left to their own devices, non-IT personnel could easily violate governance policies (and government regulations) by plopping customer data out in the cloud. “It’s dangerous when business units try to get around IT,” Microsoft’s Wilkin says. But like Linux, Blackberries, and iPhones, “cloud computing will allow business units to elude outdated IT policies until they can be updated,” Waldman predicts.
Other barriers to widespread enterprise adoption of cloud computing are only too familiar. “I have made zero progress in selling to top-level management of enterprises in terms of replacing a whole class of IT,” Novikoff says. “Moving to cloud computing can be seen as introducing unquantifiable risk.” Executives Novikoff speaks to grill him extensively about security, guaranteed uptime, and regulatory compliance and certifications, in that order. “The question is not just how secure is the datacenter, but how secure is the business process within the data center.”
Vendors will need a transparent mechanism for proving that their services meet data-protection standards. That could take the form of a quarterly report that shows the results of services regularly being tested against such standards as those for the payment-card industry. Hyperic, a maker of Web-monitoring software, already does something like that with performance metrics. Its free open-source systems-management tool peers inside Amazon’s cloud computing services to report metrics such as database query response times and system outages.
Even if cloud providers solve those critical issues, there are others. Right now the market is long on technology but short on services, which means that clients will need heavy-duty IT operations expertise. “Going to Amazon’s Web services is like going to the grocery store — you can bring home a lot of groceries but you still have to make dinner,” Novikoff says. And as companies move some applications, but not others, to the cloud, integration problems will arise. “No company runs just one or two business services,” explains Microsoft’s Wilkin. “I can’t put some of those services out into the cloud and expect the same level of integration as if everything was on the premises.”
Cloud computing won’t be the revolution that Internet adoption was, but the fact that it will take time to emerge shouldn’t discourage companies from exploring at least some facets now. The possibility of unwinding a decades-long investment in IT infrastructure in favor of a more flexible and nimble approach could even lead to that mythical IT/business alignment that companies have been chasing for years.
Vincent Ryan is a senior editor at CFO.
Peering through the Mist
Assessing the current reality of cloud computing
Hype: All of corporate computing will move to the cloud.
Reality: Low-priority business tasks will constitute the bulk of migration out of internal data centers.
Hype: The economics are vastly superior.
Reality: Cloud computing is not yet more efficient than the best enterprise IT departments.
Hype: Mainstream enterprises are using it.
Reality: Most current users are Web 2.0–type companies — early adopters.
Hype: It will drive IT capital expenditures to zero.
Reality: It can reduce the start-up costs (particularly hardware) for new companies and projects.
Hype: It will result in an IT infrastructure that a business unit can provision with a credit card.
Reality: It still requires a savvy IT administrator, developer, or both.
Cloud Computing Is:
- Standardized IT hardware, software, and other products/services
- Accessible via Internet protocols from any computer
- Always available (in theory), and will scale automatically to adjust to demand
- Pay-per-use (or advertising-based)
- Virtualized; that is, it allows customers to provision, manage, and terminate services themselves
Source: Forrester Research