Private clouds hold the promise of rapid workload deployment and dynamic scalability, allowing an organization's business units and users to create and provision new applications while maintaining full control over the data center. But private clouds also carry overhead expenses related to virtualization and private cloud software, not to mention the IT costs that can arise from poor business planning. Fortunately, organizations that...
understand private cloud costs can make technical and business decisions that will save money and improve private cloud use.
Trimming the cost of hypervisors
Every private cloud depends on a virtualized data center, and virtualization requires the use of at least one hypervisor to allow flexibility in workload provisioning, migration and protection. Hypervisors are often another layer of expense on the way to a private cloud, but organizations can save money by making smart hypervisor choices, spotting attractive licensing deals and deploying multiple hypervisors.
Hypervisors typically are free to try, allowing IT professionals to take time evaluating them and gaining valuable expertise in their use. But enterprises must also weigh hypervisor licensing costs. Each hypervisor vendor uses a slightly different scheme when it comes to licensing. VMware, for example, licenses vSphere per processor and requires a separate license for the vCenter Server management console, in addition to an annual support contract.
By comparison, Microsoft can supply Hyper-V alone for free, but it is typically licensed as part of Windows Server 2008 R2 or later. Many management features in Systems Center Operations Manager and Systems Center Virtual Machine Manager also come with additional costs.
Citrix provides XenServer and live migration free, but the added features of Essentials Enterprise and Essentials Platinum bundles are priced per server, without an obligation for support contracts.
Look for deals on the hypervisor and management tool bundles that best suit your business needs and growth objectives. The needs of a small to medium-sized business with several dozen servers will be substantially different from those of a large enterprise that virtualizes and maintains hundreds of servers.
Controlling private cloud costs before a deployment
When considering future growth, including implementing a private cloud, determine the cost of adding licenses later versus buying licenses up front. New licenses can easily be purchased, so there is little reason to purchase extra licenses far in advance.
Companies expecting significant server growth (perhaps as part of a merger or a second data center project) might want to consider volume-license discounts rather than waiting to purchase licenses later, when the number purchased will not be as large.
Conversely, organizations with exceptionally tight budgets might do better to purchase only the required licenses now and wait to purchase additional licenses until necessary. Be mindful that licensing costs and policies change periodically, so it's vital for an organization to follow its hypervisor vendor's current licensing guidelines. Not all vendors define processor in the same way, which can cause problems. Some vendors define every core as a processor, which means an eight-core processor would require eight licenses. Others define a processor as the complete CPU package, regardless of how many cores it has. This can represent a huge difference in costs. Just imagine a small business that deploys 20 servers running dual eight-core processors (40 CPU packages versus 320 cores). IT professionals should read the fine print to avoid unexpected sticker shock.
Finally, don't overlook the value of using multiple hypervisors to avoid some common cost pitfalls. Being comfortable with alternative hypervisors provides businesses with some leverage to negotiate licenses and support terms. A second hypervisor also helps protect the business against vendor lock-in, eliminates costly evaluation time and allows a confident deployment of the alternative hypervisor in the private cloud infrastructure.
This approach also allows for hypervisor tiering, meaning that less expensive hypervisors can be used for some workloads. Many organizations become familiar with multiple hypervisors and may support limited deployments of alternate hypervisors within the data center, within different computing tiers, or in secondary or hosted data center facilities.