Yes, you can have an elastic private cloud
I was having a discussion with a skeptical CIO the other day. His issue was that a private cloud isn’t real. Why? In contrast to the public cloud, which has unlimited capability on demand, a private cloud is limited by the size and capacity of the internal data center. While I understand this point I disagree and here is why. I don’t know of any data center that doesn’t have enough servers or capacity. In fact, if you talk to most IT managers they will quickly admit that they don’t lack physical resources. This is why there has been so much focus on server virtualization. With server virtualization, these organizations actually get rid of servers and make their IT organization more efficient.
Even when data centers are able to improve their efficiency, they still do not lack resources. What data centers lack is the organizational structure to enable provisioning of those resources in a proactive and efficient way. The converse is also true: data centers lack the ability to reclaim resources once they have been provisioned.
So, I maintain that the problem with the data center is not a lack of resources but rather the management and the automation of those resources. Imagine an organization leverages the existing physical resources in a data center by adding self-service provisioning and business process rules for allocating resources based on business need. This would mean that when developers start working on a project they are allocated the amount of resources they need – not what they want. More importantly, when the project is over, those resources are returned to the pool.
This, of course, does not work for every application and every workload in the data center. There are applications that are highly specialized and are not going to benefit from automation. However, there indeed can increasingly large aspects of computing that can be transformed in the private cloud environment based on truly tuning workloads and resources to make the private cloud as elastic as what we think of as a ever expanding public cloud.
Judith Hurwitz's Cloud-Centric Weblog 
I agree with you 100%. Optimization of excess resources was the foundation of EDS’ (now, HP) business model. The management of resources (either scarce or abundant – both have their challenges) is the real issue; the private cloud is about utilization, not capacity. Arguing that the “private cloud isn’t real” sounds like an argument from someone who doesn’t want to accept (or manage) a new reality. The flexibility of cloud computing is undeniable – whether private or public. Your friend may find the risks associated with virtual capacity to be untenable, and I respect that point of view. Yet balancing risk vs. reward is key to managing resources…I’m not sure that denying capabilities is the best approach to harnessing the power of the cloud.
You make an excellent point with your comment, “I don’t know of any data center that doesn’t have enough servers or capacity.” But the question always comes back to, “What is your ability to make use of that capacity?” Put another way, will your architecture let you re-deploy underused assets to solve real-time issues?
A flexible architecture needs to be an underpinning of a private cloud. Otherwise it’s just a traditional silo-oriented data center that happens to be running virtualized servers.
We faced the same issue with storage. Making efficient use of storage required a new physical architecture: the SAN. Similarly, making efficient use of compute resources requires a new architecture there.
Converging the I/O to simplify connectivity is part of the answer. The second part is the virtual I/O component that manages the connectivity.
Put both pieces together — the converged networks and the virtual I/O — and you have a great start on the elastic private cloud you describe.