Realizing the Promise of Next-Generation Data Centers

By Ivan Casanova
published: Friday, September 26 2008

As corporations are under mounting pressure to tighten their belts, they are taking a long, hard look at the way they provision applications. Traditionally, IT allocates a static amount of computing resources to fulfill user needs for a particular application at peak workloads. Essentially they are paying for the worst case scenario: the "peak" rarely occurs, so the business ends up paying for server space to sit idle the majority of the time. This means more unnecessary hardware, software, cooling and support costs. When you consider the hundreds or thousands of applications being supported, the wasted dollars spent grows out of control.

Moreover, the mere task of provisioning applications has its own inherent costs. There are many ad-hoc tasks involved with moving an application from development into the data center. Packaging, provisioning and activating programs involves numerous manual steps, such as copying folders, moving data, zipping files and more. Typically, adding a single node in a data center can take days. In fact, as much as 80 percent of an application's development effort is consumed by preparing the environment-configuring an application and changing the application environment. The effect of these fragmented processes is constrained agility in delivering application services to meet the needs of the business. Additionally, the application dependencies in today's data centers add further cost and complexity.

Shifting from Static to Dynamic Provisioning

Static provisioning results in application silos that are rigid and difficult to scale. Clearly this static approach is not sustainable. Business demand is dynamic - and often unpredictable. Either organizations end up over provisioning and have costly resources sitting idle, or they under-provision and run the risk of failing to fulfill service commitments.  Given this, they err on the safe side by setting aside computing resources for an application by over-provisioning.

To improve application service levels while reducing costs, businesses need a dynamic means of application service management. Today's organizations need to be building an on-demand infrastructure that is flexible enough to withstand unprecedented and unplanned-for levels of pressure. This allows them to gain "always-on, always-responsive" application services. While they are at it, IT organizations should take a close look at the lengthy steps involved with deploying these applications-there must be an easier way.

A New Approach

Many businesses are excited about the prospect of dynamic provisioning. The premise is simple: Whatever application has the greatest demand at a given time will get allocated the right amount of servers based on the number of requests at that time. Resources are automatically shifted as needs change.

Instead of dedicating vast amounts of computing resources to support peak usage rates, organizations can shift to a use-as-you-go approach. Services are unbundled from dedicated resources. Resources are allocated as needed to enterprise applications. This is easily implemented via a dynamic application service management (DASM) platform that essentially allows users to scale to meet fluctuations in demand. What used to be individual silos of computing resources are brought together as a virtual resource pool. This reduces the overall number of servers needed while increasing the utilization of the now smaller resource pool.

IT managers appreciate this approach because of the ability to guarantee service level agreements. By applying a policy-based approach, organizations ensure service levels while driving down IT costs. They are now able to build service-level agreements (SLAs) around throughput, latency and resource utilization, as well as a host of other criteria that was not previously measurable. Any statistics that the application service management platform exposes can be incorporated into a policy. This means organizations can, for example, right-size their pool of resources based on response times or page-rendering times. As a result, IT organizations save an estimated 20 percent on infrastructure costs without an impact on SLAs.