Virtualization to cloud: Planning, executing a private cloud migration
A comprehensive collection of articles, videos and more, hand-picked by our editors
Cloud computing is a new model of IT that's still riddled with definitions that are at best inconclusive and at
worst contradictory. One of the most fundamental questions in cloud computing is where the cloud really is.
The goal of an enterprise is to run applications, not just build expensive IT infrastructure like data centers.
The whole notion of the cloud started with public cloud resources where IT was outsourced. Enterprises involved in real cloud projects quickly realized that most IT wouldn't be outsourced, so does that mean they'll have no cloud at all, or that their cloud is private? And if it's a private cloud, what data center changes must take place?
Most enterprises go into a cloud project presuming that a private cloud is an enterprise data center architecture that, in some way, replicates the data centers of public cloud providers. When asked the question, "What service does a private cloud provide?" IT managers tend to answer that it's Infrastructure as a Service (IaaS). They see private clouds being built largely on virtualization technology. Most have no specific answer if asked how a private cloud differs from a data center that installed virtualization for server consolidation.
Unfortunately, many cloud vendors have supported this fallacy. Nearly all announcements about building private clouds are actually about enhanced virtualization tools and techniques. In most cases, the products add centralized resource management and addressing to a virtualization-equipped data center.
Some enterprises also gain early awareness of open-source cloud development tools like Hadoop or Eucalyptus. Hadoop creates a type of data model-driven cloud architecture; Eucalyptus almost recreates a virtual machine cloud similar to Amazon's EC2. If building a private cloud means building a cloud in an explicit sense, then these tools also seem to offer a logical starting point.
Justifying the private cloud
As logical as either of these private cloud visions may be, enterprises are finding that they collide with their current practices -- and come with some economic limitations.
First, many critical applications don't fit these models, and you can't easily or efficiently make them fit. This is because most of today's mission-critical applications run on multitasking computer systems, often based on principles of service-oriented architecture (SOA) on a Platform as a Service (PaaS) vision of the cloud. Virtualization isn't used or even valuable for these applications, and cloud software tools don't create PaaS clouds.
Second, the benefits of an enterprise adopting a specific private cloud model versus its current IT infrastructure may be impossible to prove, because it may not be there. Justifying the creation of a private cloud in the data center is the limited potential for gain in overall efficiency and cost.
The efficiency of a cloud-ready data center doesn't rise exponentially or even linearly with the size of the resource pool it creates; it rises quickly at first, tapers off and then plateaus. Enterprises reviewing the cost and benefits of private cloud tools are learning that their current data centers are near that plateau. Moving to an explicit cloud could increase their operations costs by adding cloud management tasks to the current operations load.
If current multitasking data center applications are cloud-efficient and are based on a SOA PaaS model like a cloud, then you might already have a cloud, for all practical purposes. And that's what enterprises increasingly believe. The goal of an enterprise is to run applications, not just build expensive IT infrastructure like data centers. If a cloud acts like an IT black box in which details are hidden, then it's fair to say most data centers are already using the cloud model of application delivery.
This new "my data center is a cloud" vision is helpful; it focuses private cloud technology planning where it should be -- ensuring the current data center structure that's delivering applications as a service is doing so with high efficiency. If not, then the enterprise can consider to use more cloud tools such as Hadoop, Eucalyptus or vCloud to increase efficiency or choose to cloudsource some applications to a public cloud.
Moving to a private cloud is necessary where data center resource utilization is low. What enterprises often find is that disorderly planning in server consolidation throughout virtualization creates inefficient resource pools that can see improvement through the use of private cloud tools.
If an enterprise hasn't consolidated servers, it's more likely that a specific application is inefficient and could be a good candidate to migrate to a public cloud such as a PaaS that is based on middleware common with the enterprise's data center. Treating a data center like a private cloud is most beneficial when planning for hybrid cloud applications.
ABOUT THE AUTHOR
Tom Nolle is president of CIMI Corporation, a strategic consulting firm specializing in telecommunications and data communications since 1982.