Today, economics can be the determining factor in adopting server-based virtualization. In other words, you can reduce server hardware costs by looking at commodity hardware. But it makes little sense, for example, to consolidate 10 virtual servers onto one physical host server if that host server is going to cost you five times more than each of the physical servers you plan to phase out.
This might not seem like an issue for organizations considering blade servers, given the general perception that they are an inexpensive option. (It’s true that blade servers have come down in price and are increasingly attractive because of it. But a basic chassis will eventually require add-ons for it to serve an organization’s needs, and in order to do that, it will cost more.)
Are blade servers all that sharp?
The problem is that unlike a rack server or tower servers, blade servers are not self-sufficient. They are small because they do not include resources such as power supplies or fans. These components and several others are instead integrated into an expensive chassis, which serves as a backplane to the blade servers. This provides them with power, cooling, network ports and various interconnections.
Organizations must consider the cost of the chassis and the cost of any necessary add-on modules when determining the overall costs of deploying a blade server. As one example, if an organization needs only a few servers, then the per-server cost of blades will be much higher than for comparable rack servers. The cost of the chassis should also be factored in along with any required support modules. On the other hand, if you completely fill the chassis with all the blade servers it can accommodate, then the price per server comes down significantly because the cost of the chassis is spread across many servers.
Bottom line: If an organization is looking to save money by reducing the number of physical servers purchased, blade servers may not be the best choice. They are only cost-effective when purchased in quantity.
One other element favors rackmount servers in virtualized data centers: Lower-end blade server configurations often lack sufficient hardware to host large numbers of virtual machines (VMs). If your larger organization plans to deploy hundreds of virtual machines, you will have to factor in the cost of more physical blade servers.
A license to operate and related services
Keep licensing in mind when you consider costs for both hardware and Windows Sever 2008 R2, particularly for the Enterprise Edition. Under Microsoft’s current licensing agreement, you can run only one instance of the operating system in a physical environment and up to four instances in virtual environments. But if you operate all five instances simultaneously, the instance of the server software running on the physical operating system may be used only to run hardware virtualization software and to manage environments on the licensed server. You are allowed to use only the one running in the physical environment to manage OSes on the licensed server. If you need to run additional virtualized servers, you must purchase licenses for additional physical servers, which can add up.
Yet another important element to think about is the sort of maintenance and services contract you should purchase and from whom. If your Windows server is hosting your online business and you experience a hardware failure that brings down your entire operation, you will need a contract that guarantees a one- or two-hour response time instead of one that promises 12 hours.
It is also a good idea to specifically stipulate the number of hot spares a service provider should keep on hand for you. Obviously, the faster the response time, along with a number of other services you need, the more expensive the contract.
If a third party hosts your business and you also want the company to maintain your hardware, consider uptime guarantees. Many providers offer guarantees that their network will be up 99.99% of time (except for short periods involving scheduled maintenance). But whether four 9s reliability is enough or whether you need to go to five 9s (which would be just five minutes of downtime a year) or higher is a decision to be weighed against what your own IT shop can provide compared to a service provider’s offerings.
For instance, if you plan on moving a major portion of your internal and external business to the cloud, you may want the security of having five 9s or higher reliability. But achieving this level of reliability requires multi-million dollar investments in redundant infrastructure.
Interestingly, most IT shops decide to purchase service and maintenance agreements through third parties rather than the vendor they bought the hardware from. The two top reasons for that choice are cost and proximity of the service company to their business.
Ed Scannell is Senior Executive Editor with SearchWindowsServer.com. He can be contacted at firstname.lastname@example.org.