Open-source cloud frameworks: A work in progress

Good fits for open-source frameworks tend to include experimental cloud applications built by developers who are comfortable with newer, open-source tools. Other likely candidates include applications deployed by organizations such as universities or research labs, which have the technical skills to learn and work with these new technologies, and/or the need for specialized capabilities such as massive databases or advanced analytics, says Roby.

Typical apps deployed using open-source frameworks include Web and social applications, as well as mobile or customer-facing websites, says Jerry Chen, vice president of cloud and application services at Cloud Foundry. Such frameworks are also useful when organizations need to deploy applications quickly and scale them up and down as needed.

Legacy applications requiring hardware or software that may not be supported on the Web tend to be less attractive candidates. "While it is very possible to migrate many data center applications from local servers onto [virtual] cloud-based ones, the ROIROI is not always clear," says Bill Weinberg, senior director of Olliance Group at software and services provider Black Duck Software. "The downside can lie in potential security issues, divergent response to loading, throughput bottlenecks and availability." Alles zu ROI auf CIO.de

OpenStack and Cloudscale are better choices for complex applications than Eucalyptus, says Nataraj, because they do a better job of hiding the complexity of networking. For an application that, for example, requires a user "to connect from a different IP range," a customer would "have to write custom code to make that happen with Eucalyptus," he says. With OpenStack, the "switches" required to make those new network connections are already present.

The number and quality of developers involved in an open-source project can also be a good indication of the project's quality, many observers say. If developers from several companies are involved, vendor lock-in is less likely to be a problem, says Nataraj.