One of the great tenets of cloud technology is that it scales. Whether you’re dealing with virtualized servers or storage, when it’s time for more resources you just go out and consume them — if you’re using the cloud, that is. This makes public clouds a great place to hatch new ideas: spin up some resources, push some data around and share your creation with the world. If it takes off, you can add more storage, instances or bandwidth. If it doesn’t, you can shut everything down and move on to plan B.

But what happens when your application really takes off? When you need to be pushing around petabytes of data instead of terabytes? When dozens of instances becomes thousands? Or when you need to put a faster network between your application and your database? What happens when the auditor comes calling and you need to ensure that you’re complying with PCI/HIPAA/FISMA requirements? What do you do when you need to put an application presence somewhere in the world where there are no public clouds? For any or all of these possibilities there may well come a day when you need to consider not just consuming someone’s cloud technology, but building your own cloud.

Much like making the leap from renting to owning a home, owning your own cloud comes with its own set of new responsibilities. You have to mow the grass and fix the leaky faucets. But, you also get the opportunity to put in custom countertops and install a smartphone-controlled garage door opener. When designing a private cloud of your own, you can customize to your heart’s (or budget’s) content. This gives you a broad range of flexibility, but also introduces an interesting concept to folks accustomed to consuming someone else’s cloud –clouds run on hardware. Making the right choices for the servers and network equipment that power a private cloud can net huge performance gains for the applications that live there.

When you’re renting, there’s always a bigger apartment or bigger house. One with a pool or a great view of the city’s skyline. Once you own a home, however, adding on to it requires a bit more of an investment. The same is true of a private cloud. With public cloud technologies you typically view your ability to expand as being limited by things like API rate limiting. With a private cloud, however, the major limit is going to be based on the number of physical hosts you put into your environment. You need to have a strong understanding of what your short and near term growth look like to ensure that you have appropriate physical resources available to scale.

Owning a cloud can have benefits beyond being able to customize the servers and network equipment that you use. It also provides you the option of becoming a landlord. If you build extra capacity into your private cloud you have the option of reselling those resources. While you may not want to let the general public onto your infrastructure, there may be other projects or departments within your own organization that could take advantage of the cloud resources to help them achieve many of the same goals you set forth when you decided to build your own cloud.

The CTO of Zynga has stated, “We realized we needed to own the base and rent the spike.” This concept describes the reality many large-scale cloud consumers are moving toward: investing in the infrastructure necessary to support the baseline volume of traffic your application receives (with a little headroom for growth) and then bursting into public cloud infrastructure when necessary can give you the best of both worlds. For some workloads, this model is a reality today. Over the next year or two, advances in network virtualization and the integration between public and private cloud platforms will significantly expand the use cases for this model. Using a blend of public and private cloud resources creates an additional “jet setter” persona that is akin to owning a home and renting condo/cabin/beach house for an extended getaway.

The options available for launching new ideas in the cloud continue to evolve and expand. This encourages experimentation and innovation, and allows for the best of those ideas to mature and grow. Enterprises that already have large footprints can experiment with new ideas far more easily than they could have five years ago. This creates an incredible ecosystem where an apartment dweller can become a homeowner or a jet setter can decide they would be better off just paying rent.

Find out more about the differences between virtualization and cloud computing, check out this white paper that Jacob contributed to.

Virtualization is an excellent option if you want to consolidate a hardware footprint, increase service availability or move legacy applications onto modern hardware. Virtualization will allow you to move your existing application architecture onto updated hardware without having to rethink how your services are deployed.

Do you have two web servers and a database cluster with SAN storage? Then spin up four virtual machines (VMs) on two physical servers. The solution will still look the same to your end users, but you can achieve higher utilization of your hardware by allocating the appropriate resources to the individual VMs that need them. Add a third server and you can take advantage of failover technologies that will allow you to move virtual machines between physical hosts and protect you from hardware failures.

For workloads that are largely static, you can build virtualization infrastructures out to be fairly massive in scale. There are well-established tools available to simplify management of hundreds or thousands of virtual machines. Virtualization allows you to save money on hardware infrastructure, but does not significantly impact the labor involved in managing the machines.

Cloud based solutions differ from virtualization solutions when you start looking at dynamic workloads. Rather than managing hundreds of VMs, an IT organization can simply allocate resources to various customer groups. If an individual cloud instance starts returning bad data it can be destroyed and replaced with a freshly provisioned one. If you need more resources dedicated to serving a public-facing website or handling report generation at different times of the day, tools exist to reallocate your pool of resources.

Moving an application from traditional IT infrastructure to the cloud typically does require you to rethink how the application itself is deployed. This can require an investment in retooling up front, but over time will allow you to save money both on hardware infrastructure and labor.

Virtualization is a tried and true technology and is still a good fit for many workloads today. If you’re looking to improve redundancy, become more resilient to hardware failures or move aging application environments onto more modern hardware, virtualizing the servers and taking advantage of automatic failover technologies is an excellent path to travel. However, for solutions that require massive scalability, workloads that can be readily parallelized, or situations that call for an increasing level of consumer self-service, cloud solutions offer very clear advantages.