Today,server sprawl is an insidious problem facing most midsized to large organizations. When a company acquires or creates new applications, typically the organization deploys each new application on its own server. On the surface, small-scale server deployment for departmental or branch offices to support new applications seems inexpensive. However, each new application equals a new server.
Functionally, deploying each application on its own server works OK, but financially, there are many downstream costs that businesses might not be aware of. For example, many businesses pay for system maintenance, which covers the repair or replacement of defective systems. Many businesses also implement a hardware-replacement policy that states all hardware must be refreshed every 3 to 5 years.
Added servers multiply these costs, and server sprawl escalates them. As a business deploys additional servers, the business must plan for costs involved in replacing those servers down the road. Other not-so-obvious costs include increased infrastructure costs. The more servers a company has, the higher the electrical and cooling costs—not to mention increased requirements for floor space and networking.
Another problem that develops out of server sprawl is hardware utilization. Most systems that rely on this deployment style have utilization levels ranging from 5 to 20 percent. This can be good for response time but it also means the business isn't getting its full value from its hardware purchases.
The real costs of server sprawl lie in the administration—each penny spent on hardware costs a dollar to manage. IT efficiencies drop as the number and types of server platforms proliferate. Most organizations utilize servers from multiple vendors and rarely standardize on one OS, so IT ends up dealing with a plethora of hardware and OSs. Different hardware systems have different support requirements and characteristics,which means IT personnel must have many different skill sets. Plus, different OSs typically require different patch-management strategies as well as different backup and restore procedures. These types of demands reduce IT efficiency and increase costs of ownership.
Server consolidation provides a solution to server sprawl by combining the workloads running on multiple hardware platforms. Server consolidation comes in three major types: workload partitioning, hardware partitioning, and server virtualization. Workload partitioning attempts to combine similar workloads but can be hindered by application incompatibility and complex workload analysis. Hardware partitioning can be a good alternative but requires high-end (i.e., usually Itanium-based and expensive) hardware, and the number of partitions even high-end systems support is limited.
In most cases, server virtualization is the best way to implement server consolidation. Virtualization doesn't require specialized hardware, and it lets you run multiple OSs with complete application integrity. Better yet, two of the major virtualization products, Microsoft Virtual Server 2005 Release 2 and VMware Virtual Server 1.0, are both production ready and completely free. At the high-end of the virtualization market, VMware ESX Server provides even higher levels of control and scalability—but it's not free. If your organization is facing server-sprawl problems, you might find that today's virtual technologies can provide real solutions.
