eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
The future of computing may be a lot harder to predict than the weather. So, can you bet your company on the cloud?
This question is pressing on enterprise IT executives who are trying to weigh the allure of low-cost, flexible computing in the cloud against the risks inherent in the platform. The fear of outages is perhaps one of the biggest obstacles to the adoption of cloud computing, but IT and business managers also worry-rightly-about data security, regulatory concerns and vendor viability, among other things.
Interest in the sometimes nebulous concept is rising faster than moist air in a thunderhead. Venture capital companies are showering investment dollars on a crop of tiny, oddly named startups. Established vendors are scrambling to reposition their offerings as “cloud this” and “cloud that.” Dell has even sought-unsuccessfully so far-to trademark the expression “cloud computing.”
Definitions vary, but a few generally agreed-upon characteristics of cloud computing are emerging: It is not a product but a service; it is available via the Web with little or no human assistance; it is more or less instantly and infinitely scalable to accommodate sudden bursts and drop-offs in demand; and customers pay according to usage with little long-term commitment.
Amazon Web Services, or AWS, are generally agreed to be quintessential cloud services. Amazon.com’s EC2 (Elastic Compute Cloud) for application development and its S3 (Simple Storage Service) both serve up a Web interface, inviting the customer to pay for what’s used via credit card.
Amazon.com has augmented those services with SimpleDB, Simple Queue Service, Flexible Payments Service and Mechanical Turk, a service, now in beta, that provides an on-demand work force.
In August 2008, Amazon.com enhanced EC2 with EBS (Elastic Block Storage), which enables storage to persist after an EC2 instance is terminated.
Amazon.com is far from alone. Search engine giant Google offers Google App Engine, a cloud-based platform for application development. In the United Kingdom, hosting provider XCalibre Communications is serving up FlexiScale, which offers users a self-service virtual dedicated server on the Web.
“You go online and sign up,” said Philipp Huber, chief operating officer of XCalibre. “Once you have an account, you can set up a machine in a minute or two.”
Most observers agree there are several basic types of cloud. Forrester Research has segmented the cloud market into five categories: Web-based services and SAAS (software as a service); developer application services; middleware infrastructure as a service; virtual IT infrastructure as a service; and physical infrastructure as a service.
But not all clouds are created by commercial cloud computing service providers. Most significantly for corporate IT professionals, corporate clouds-call them “intraclouds”-have emerged, in which companies with significant IT resources make them available in cloud form to their own employees.
And in what may be a sign of the staying power of compute clouds, a cloud ecosystem is developing in which utility vendors are creating tools for monitoring cloud performance, and cloud aggregators are putting together packages of different cloud services.
Risk vs. Reward
Despite generating widespread enthusiasm, cloud computing faces the same hard questions that remote computing services have always faced. As with cloud computing precursors ASPs (application service providers), SAAS vendors and the like, customers want to know if their data is secure, whether they’ll be compensated for outages, what happens to their data should their cloud provider go under and whether their data may be going offshore.
Periodic outages are a signal to enterprises to be wary. AWS suffered 8 hours of S3 downtime in July, echoing an earlier outage of several hours in February. In August, Google Gmail and Google Apps messaging and collaboration cloud services suffered an outage of nearly 15 hours.
“There are some problems still there-where the data resides, for example,” said Gartner analyst David Smith. “We are early in the cycle-those who are evaluating [cloud computing] are asking those questions.”
Forrester analyst James Staten notes that companies may not be able to certify the integrity of their data for regulatory purposes if they do not know where it is while it’s being processed in a cloud-computing service.
For TC3 Health, a health care claims processing and cost containment company, the cloud payoff was well worth the risk. TC3 needed extra capacity when a customer asked to have a large number of claims checked for accuracy.
“We had 30 million claims dumped in our lap. We were overloaded,” said Paul Horvath, CTO at TC3 Health.
Using TC3’s standard processing approach would have cost $750,000. Instead, TC3 tapped AWS for additional capacity, at a total cost of $220,000, including a fee paid to RightScale, a company that customizes Amazon.com’s cloud services for clients.
“I pay by credit card,” said Horvath. “The cost ranges from a few hundred to several thousand per month.”
TC3 has six different server configurations on AWS, ranging in cost from 10 cents per hour per server to 80 cents per hour per server.
With regard to security and confidentiality, Horvath said, confidential information is stripped out of the forms before they’re processed on AWS, and data is encrypted while being fed into and retrieved from AWS servers. Although Horvath does not know the exact location of the servers, he said he understands that they are in the United States.
Although some mainstream businesses such as TC3 are reaping the benefits of commercial cloud services, the principal cloud beneficiaries so far have been small companies that can’t afford extensive IT operations and IT staff-the same companies that were targeted by the ASPs of a decade ago. In many cases, cloud-computing startups are serving other startups.
Rick Wee, founder and CEO of Blissbook.com, a social network for the bridal community that is about to launch a beta site, selected GoGrid, a cloud service that is itself currently in beta, to provide a development platform for the site. Wee had launched several previous startups and, in doing so, felt the heavy burden of investing scarce funds in IT infrastructure.
“Cloud computing reduces startup costs substantially,” Wee said. “You pay for what you use. We recognize the need to grow business in a methodical manner, not overinvesting or underinvesting, but able to scale quickly without a large infrastructure investment.”
Some cloud providers, such as GoGrid, are enabling application development with Ruby on Rails, a streamlined version of the Ruby programming language that is popular among social network sites, according to Wee.
Similarly, MLB.com, the Web site of Major League Baseball, sought to accommodate sudden surges in use that typically occur, for example, at the time of the World Series and the All-Star Game and when traditional rivals meet.
A year ago, MLB.com launched a chat application consisting of 100 chat rooms per major league game. That’s plenty for most games, but not when, for example, the Boston Red Sox play their archrivals, the New York Yankees, said Christian Gough, system administrator for MLB.com.
“We’re using Joyent, which allows us to meet spikes in demand,” said Ross Paul, architect of enterprise service network MLB.com, a wholly owned subsidiary of MLB.
Founded in 2004, Joyent provides several cloud services, including a Sun OpenSolaris infrastructure for running Web sites.
“I don’t have to build new servers, rack them up and wire them up in our data center,” Gough said. “I can call Joyent at any time and, in a few hours or less, can provision the servers and get running.”
“Joyent can provision any type of server with a whole software stack,” added Paul. “It completely removes the overhead and administrative cost.”
Having successfully used Joyent to backstop the chat application, MLB.com is aiming to use Joyent services to handle other applications, such as one that’s under development on Facebook.
The Intracloud
Although the use of clouds to enable social networks adds a coolness factor that helps fuel industry hype, clouds are no less beneficial at large enterprises and government agencies, which are building clouds for the use of their own employees.
The Department of Defense’s Defense Information Systems Agency built RACE (Rapid Access Computing Environment), a cloud of computing resources for use by DOD personnel.
Alfred Rivera, director of computing services for DISA, said commercial clouds such as Amazon.com’s EC2 inspired the DOD cloud, which was created with the help of Hewlett-Packard. RACE, which Rivera likens to the PAAS (platform as a service) model, can be used not only by support staff but also by combat forces and those responding to disasters.
“When Katrina hit, we had to send forces,” Rivera said. “It would have been nice to have been able to provision an OS environment on the cloud where the first responders could have pulled the necessary data.”
To use RACE, DOD “customers” swipe a card and pay by the month for an environment that can be provisioned within 24 hours. A customer can increase the capacity at any time; if it is unused, it will be turned off at the end of a month. A charge of $500 per month covers a VMware-based Windows or LAMP (Linux, Apache, MySQL, Perl/PHP) virtual environment.
Although a suite of applications will be available when RACE goes live in October, users can deploy their own applications in their cloud environment, Rivera said.
The Big Players React
Just as HP supplied core technology for RACE, IBM sees such corporate intraclouds as fertile territory on which to market its ability to provision and run data centers.
“IBM is on a parallel strategy to Amazon and Google,” said Irving Wladawsky-Berger, chairman emeritus of the IBM Academy of Technology. “IBM is in the business of helping our customers so they can offer cloud services to their customers.”
One variation on this theme is a cloud-computing center that began operation in May at a software park in Wuxi, China, 2 hours west of Shanghai. IBM provided technology, including System x and System p servers on a secure VLAN (virtual LAN) for the center, which is run by the Chinese government and will serve as a cloud-computing resource for software companies operating in the park.
Providing IT services as a utility enables small companies to operate without the burden of upfront costs. “In an emerging economy, there are fixed costs to start up a business,” said Dennis Quan, IBM’s CTO for high-performance on-demand solutions. Flying under the Blue Cloud banner, IBM has also built data centers to offer cloud-computing services in Beijing; Dublin, Ireland; and Johannesburg, South Africa.
Dell, similarly, takes a data-center-centric view of clouds that serve a company’s internal needs. “We’re really seeing that,” said Todd Brannon, market development manager for Dell. “That’s an evolution of HPCC [high-performance cluster computing]. Maybe during the day, you’re doing thin client; at night you’re doing number crunching.”
Brannon said Dell is providing cloud infrastructure to two large social media sites, although he declined to name them. Seeking to score a marketing coup, Dell also filed a trademark application for the term “cloud computing,” a request that was denied in a nonfinal judgment in August.
Meanwhile, all eyes will be on Microsoft at its Professional Developers Conference in October, when the company is expected to make a major cloud announcement. Microsoft already offers SharePoint and Exchange hosted services. A company spokesperson declined to comment or to make Microsoft executives available for this article.
The Enablers
While the big players are busy adapting their skills and cranking up their marketing machines to address the cloud paradigm, small, agile companies are scrambling to provide utilities and services to cloud customers. In the case of TC3, RightScale helped the company create a scalable application to run on Amazon.com’s EC3.
RightScale also offers a dashboard to enable users to set up, launch and monitor AWS cloud services.
With reliability sometimes being a problem, cloud service monitoring tools are taking on an important role.
Hyperic, a company that sells tools for managing large-scale Web environments, recently introduced the beta of CloudStatus, a tool for tracking AWS service levels.
“Cloud platforms are great, but you need the monitoring and management you have with on-premises infrastructure,” said Javier Soltero, CEO of Hyperic.
Soltero added that there is a lag between the time of an AWS outage and when Amazon.com reports it. In the S3 outage that AWS suffered recently, he said, CloudStatus was notifying its users via Twitter of the outage 20 minutes ahead of Amazon.
Two vendors are offering tools to integrate cloud services.
Kaavo is a startup offering IMOD (Infrastructure and Middleware on Demand), now in beta, to integrate different AWS EC2 and S3 services from a browser.
Boomi, a company that specializes in integrating applications for small and midsize businesses, recently introduced Boomi On Demand, which lets customers integrate different combinations of SAAS offerings from a Web interface.
Just as the rapid emergence of applications and utilities inaugurated the era of the PC, the rapid emergence of such tools in a cloud ecosystem augurs well for the staying power of the cloud phenomenon. But that may not be enough to get large enterprises to lean heavily on commercial clouds, even as they build out their own intraclouds.
“Private clouds will be huge; the big guys will not trust public clouds for a long time,” XCalibre’s Huber said.
But intraclouds are part of the cloud picture, as well. And when you factor in the large companies’ market muscle and the stimulating jolt of venture capital dollars, not to mention the proliferation of cloud-computing providers themselves, it adds up to quite a lot-enough, even, for the next big thing.