Private clouds seem like a great idea. They offer the flexibility, scalability and self-service capabilities of...
public cloud, but with the control of a private data center. But despite their potential, private cloud deployments are complex and demanding -- and sometimes, they fail. Many organizations dream of Amazon-like capabilities, but are unprepared for the technical challenges.
At Gartner's IT Operations Strategies and Solutions Summit last month, research director Mark Lockwood explained how unrealized benefits, uninterested users, poor return on investment and other issues threaten private cloud success. Here are eight ways a private cloud project can stumble.
Politics and team structures
Many companies patch together a cloud team from various IT silos. However, other projects, company politics and varying levels of commitment among team members can cause big problems in a private cloud project. It's difficult for disparate IT personnel to work together seamlessly for a common goal. The result is a private cloud that's poorly conceived, designed, deployed and managed.
Poor process or governance
Private clouds build on well-established virtualization environments. However, without creating new processes, especially for automation and governance, organizations will struggle with private cloud control. "A self-service portal needs governance to stop 7,000 new VMs from being created in a week," Lockwood said.
Excessive automation complexity
Automation is an essential component of private cloud services. But automation can be complex, invoking large numbers of detailed and dynamic rule sets. To create these rule sets, organizations require detailed knowledge of workflows and processes that may not have existed in a pre-cloud world. IT teams also need a keen knowledge of automation tools to translate processes into rules. What's more, they must periodically review and update the rule sets, since automation is not a "set-and-forget" activity.
Inadequate third-party integration
Private clouds are not ubiquitous -- successful operation depends on smooth interaction between automation tools, management and reporting platforms, as well as other technologies, such as OpenStack. To avoid integration problems, organizations should support directory services and expose APIs.
No long-term commitment
Developing, deploying and maintaining a private cloud is not a quick or simple endeavor. Many of the technologies that support a private cloud cannot be changed easily. For example, moving from vSphere to Hyper-V, changing resource management tools and swapping automation platforms can be extremely disruptive to a cloud environment. This means organizations need to select platforms carefully and with an eye toward long-term use. Organizations that haven't standardized on a hypervisor or use multiple management tools may not be ready for private cloud. And if integration issues arise after deployment, the business may be forced to live with disruptions for a long time.
No public cloud compatibility
Private clouds can exist alone, but rarely do. Most organizations lack the infrastructure to accommodate a huge number of instances or cloud applications with erratic or highly scalable resource demands. This often leads to a hybrid cloud strategy, where private cloud services are integrated with public cloud services from providers such as Amazon Web Services (AWS) or Google. The hybrid cloud model is especially useful for tasks such as big data processing or cloud bursting.
Inadequate team skills
Private cloud deployments require more than just provisioning VMs and storage. They require skill sets around orchestration, automation, lifecycle management and more. That's a tall order for any IT staff, especially those that are new to virtualization or automation. This makes it particularly important to formulate the right team from the start of any private cloud project. Organizations should develop or augment IT skills as needed, or even add new staff members. Otherwise, the private cloud project will risk delays and costly oversights.
Lastly, private clouds require adequate underlying infrastructure, including hypervisors, storage and network connectivity. As organizations add new platforms to support private cloud, data center infrastructure upgrades and expansions may be required to ensure adequate reliability and performance.
Small steps, quick wins with private cloud
While the odds might seem stacked against a private cloud project, there are ways to avoid potential pitfalls.
First, abandon the idyllic notion of replicating AWS or Google -- it won't happen. Instead, focus on the top private cloud benefits for your organization and outline the goals you are trying to achieve. For example, if you need a private cloud to support self-service provisioning of test and development instances while enforcing quotas and lowering costs, start there and build out other functionality later.
Next, implement the right process and organizational structure to support your private cloud. Determine the processes needed to make the cloud work, and make sure there is a team to own and support it. Then, to achieve automation, fine-tune your private cloud processes.
"Automating a broken process won't fix it," Lockwood said.
Once you decide what the private cloud should accomplish and the processes needed to support it, implement an appropriate level of automation to meet immediate goals. Avoid overly complex or ambitious automation projects. Complexity is the mortal enemy of a private cloud project, so implement enough automation to support the most frequently provisioned workloads. More automation can be added as cloud features develop.
Private clouds can bring enormous flexibility and efficiency to an enterprise, but their success isn't a guarantee. When organizations set overly ambitious goals, or try to replicate major public cloud providers, private cloud projects can fail. Approach private cloud deployments with the goal of achieving smaller, faster wins that bring tangible benefits to the business. Then, build out functionality as users adopt the private cloud and realize its capabilities.
About the author
Stephen J. Bigelow is the senior technology editor of the Data Center and Virtualization Media Group. He can be reached at email@example.com.
Planning a private cloud migration
Five ways to achieve private cloud automation
Private cloud benefits not always guaranteed
Comparing top private cloud management tools
Dig Deeper on Cloud architecture design and planning