Building a Private Cloud: Bottom-Up, Not Backward
Tuesday, December 10, 2013
In the journey to build a private cloud, the way the individual components come together is critical. Designed well, you’re creating an automated, cloud-optimized data center—one that empowers the business to adapt to a fast-changing landscape. Designed poorly, and you’ve just added another layer of complexity, reinforcing IT’s role as a cost center rather than a strategic enabler.
There are two primary approaches to designing your private cloud:
Top-down, where you begin with an orchestration platform like OpenStack and layer commodity hardware beneath it;
Or bottom-up, where you build from the infrastructure you already have, selecting an orchestration layer that best fits your current environment.
For most large enterprises, the bottom-up approach tends to win. Why? Because it leverages existing investments in architecture, infrastructure, and staff expertise—typically centered around a standard hypervisor already deployed in production. That said, some organizations are ditching legacy constraints altogether, opting instead for a clean-slate, open-standards approach. They’re designing new cloud-native environments and either migrating applications or rebuilding them to thrive in this new world.
For the purposes of this article, we’re focusing on the bottom-up approach. (If you’re exploring OpenStack, check out OpenStack.org—we’ll cover that approach in a future post.)
Virtualization-Enabled Core Infrastructure
Your entrenched infrastructure can be the foundation of your cloud strategy—if it supports advanced hypervisor features. Tools like vMotion, Live Migration, Replication, vShield, Virtual Distributed Switches, and others are only available when the hardware meets certain vendor specifications.
If your compute, storage, or networking gear can’t take advantage of these features, you’re stuck paying for software capabilities you can’t use—wasting both CapEx and OpEx. Worse, you might be forced to invest in additional tools and services just to make up for hardware limitations.
Converged Infrastructure (CI)
One of the most impactful trends in private cloud is converged infrastructure—where servers, storage, and networking are managed as a unified whole. CI offers powerful benefits: single-console management, streamlined provisioning, and automation across the data center stack. Want to spin up a VM? Do it in one click, with storage, memory, and network all handled automatically.
But buyer beware: many CI offerings come with vendor lock-in. You’re not just buying a platform—you’re committing to a vendor’s ecosystem of compute, network, and storage. Essentially, it’s a return to the client-server mainframe model, just dressed in modern cloud clothes.
Ask the hard questions:
How does this solution integrate with my existing infrastructure?
What happens when it’s time to refresh hardware in three years?
Does this CI platform support open standards?
Some vendors are embracing openness. Others… not yet. Caveat emptor.
Virtualization Layer
Choosing a hypervisor is often less a question of “which is best?” and more about what your team already knows and what’s already deployed. In most enterprises, you’ll find either VMware ESXi or Microsoft Hyper-V. They’re feature-rich, deeply integrated, and well supported—but not always cheap.
If your private cloud goals include open standards and cost control, XEN and KVM stand out as strong alternatives. While they may lack some of the polish and third-party integrations of their commercial counterparts, they’re far more open and can significantly reduce licensing costs—especially at scale.
The right answer often comes down to what your team is already trained on and where your business wants to go.
Orchestration Layer: Enabling Cloud Behavior
The orchestration layer—sometimes referred to as the “cloud layer”—is what gives your private cloud its cloud-like behavior. According to NIST, there are five essential characteristics of a cloud:
On-Demand Self-Service
Broad Network Access
Resource Pooling
Rapid Elasticity
Measured Service
Your cloud doesn’t need all five to qualify—but the more you include, the more cloud-native your architecture becomes. At a minimum, a private cloud should deliver the first three. Rapid elasticity and metering may come later, but they’re essential for scaling and cost transparency in the long term.
Cloud Management: The Hybrid Imperative
Cloud management becomes crucial when your private cloud extends beyond one data center or blends with public cloud environments. A cloud management platform (CMP) allows for workload orchestration, policy enforcement, and seamless workload mobility across hybrid environments.
This is the holy grail of cloud strategy: the ability to move workloads between clouds—private or public—based on cost, performance, or compliance requirements, with minimal friction.
In Closing
Designing a private cloud isn’t about chasing trends—it’s about aligning with your infrastructure, your team’s skills, and your future goals. Whether you build from the top down or the bottom up, the only wrong choice is doing nothing.
The cloud is not a product. It’s a model—and it’s evolving. Enterprises that move with it will outpace those who don’t.