The cloud isn't an easy landscape to navigate, but CIOs can take a page out of cloud computing service providers' blueprints to get started -- and create the foundation for a next-generation architecture now, even if they don't hop on the cloud computing train.
In many ways, the cloud computing model is an evolution of what is already happening in data centers: virtualization, automation and dynamic infrastructure provisioning, all aimed at creating real-time infrastructures for flexible and dynamic service delivery.
"What cloud computing has done is put all these technologies together, like virtualization, SOA, Web apps, virtualization and grid computing, to create highly scalable and elastic IT capabilities as services," said David Cearley, an analyst at Stamford, Conn.-based Gartner Inc. "This is something that organizations should be aiming for anyway in their own data centers."
Cearley recommends that CIOs start experimenting with some low-hanging fruit like sales force automation, a tested Software as a Service model, to see how cloud providers use automation techniques and pooling and sharing resources, and then take those best practices and apply them to their own data center.
Taking it a step further, CIOs can start directing the development of applications for Web-based environments, as traditional applications are not designed to take advantage of multi-tenant or parallel processing models found in cloud computing.
"CIOs thinking strategically will re-architect applications to take advantage of Web-oriented architectures that exploit the distributed data and distributed processing model that exist in the cloud," Cearley said. "By doing so, their applications are ready to exploit the external cloud as it matures, but at the same time [the company] gains efficiencies and advances in their own environment."
The application development team is a good starting point, but CIOs need to keep in mind how the move to a public cloud would affect network traffic, the finance department's billing models, backup and disaster recovery strategies, interoperability with in-house applications, data privacy and the list goes on.
"A cloud deployment has to be a whole company decision, where you can't ignore even one domain," said Chris Howard, an analyst at Burton Group Inc. in Midvale, Utah. "It may look great from an application point of view, but the network guy has to support that chatter over a real environment, or risk management may say 'no way' to putting the data you want to outside the firewall. And CIOs run the risk of cutting corners if they rush cloud projects and have to deal with integration issues on the back end farther out."
Picking the right applications for cloud computing
Workloads or applications with unpredictable or capacity requirements play well in the cloud computing model, given that IT can spin up server and storage capacity on a cloud provider's network for a project such as a marketing campaign, and turn the capacity off once the project is over. The cost is a monthly fee, or compute capacity charges for the duration of the project versus buying your own servers that sit idle until those spikes occur.
Application development and testing in the cloud is a service that companies like IBM, Amazon.com Inc. and Skytap Inc. offer, giving customers an environment that reflects their own test and development setup. Benefits for the business can be significant: faster project times to market, given that IT is not waiting for servers to open up to start testing a new application; the ability to tell a department that the new application it wants tested will take a couple of days or weeks vs. months; and of course, the knowledge that you will not have to expand your environment or even build one.
CIOs thinking strategically will re-architect applications to take advantage of Web-oriented architectures that exploit the distributed data and distributed processing model that exist in the cloud.
David Cearley, analyst, Gartner Inc.
Autonomous applications, or ones that don't interact much with the back office, are also a good fit for cloud computing. However, if an application in the cloud relies heavily on another application or set of applications that still live in your own data center, that cost of that data transfer can be significant.
CIOs are steering clear of back-office applications in the cloud. The main reason is that cloud providers are not offering the same levels of security controls and performance assurances that organizations have in-house.
"A lot of CIOs believe that the cloud is a great way for them to get rid of their most painful applications. That's the wrong answer," said James Staten, an analyst at Forrester Research Inc. in Cambridge, Mass. "Back-office apps are not designed to scale on a cloud platform. They are not simple to administer in a straightforward way on the cloud to ensure reliability, and they often require clustering, which is tricky on any platform."
Some experts believe service-level agreements (SLAs) will prevent CIOs from putting apps out on the cloud. As it stands, robust SLAs for security, performance, reliability, availability and pricing structures that won't come back to bite you are just not in place yet.
Err on the side of caution when signing an SLA as far as restrictions for things such as traffic spikes.
"Organizations will sign a contract with the intention of using a certain amount of capacity on a regular basis with allowances for spikes in traffic up to a certain point," said Anne Thomas Manes, a Burton Group analyst. "But if the spikes go higher than what you signed up for, performance will be affected because you didn't sign up for that extra capacity."
And on the flip side, the more elasticity you want built into your cloud environment for spikes, the more it will cost you, she said.
Let us know what you think about the story; email Christina Torode, Senior News Writer.
This was first published in July 2009