March 25, 2013
As data centers are asked to support an increasing variety of applications, the desire for customers to house all of them within a facility has escalated accordingly. While this customer requirement is inherently understandable, not all data center providers offer a compatible solution. To avoid this potential misalignment between a customer’s requirements and the ability of a provider to address them dictates that customers must understand how their potential data center partners address the need for mixed load groups.
For the sake of simplicity, mixed load groups reflect varying power requirements for two or more applications. The ability to support these mixed configurations is typically not an issue for providers who use the data hall ( monolithic, traditional, and standalone) design. In these structures each data center (data hall) is physically constructed at the site itself with ample space for varying density racks.
Due to their limited physical size pre-fabricated data halls are incapable of supporting a large number of varying server racks. Correspondingly, this limitation eliminates their ability to support heterogeneous applications within a single container. Thus, any effort to support mixed load groups within a containerized environment requires that specific containers be dedicated to specific homogeneous load groups. Apart from the expense associated with having to purchase a container (or containers) to house each individual load group, this arrangement is also less than desirable from a planning standpoint as capacity can easily be stranded. Let’s use an example to illustrate the point further. Assume that a pre-fab data hall supports 200kW and has room for 40 racks. That’s 5kW per rack. If a user puts in 10 racks that only total 20kW (networking legacy, etc.), then that means that the remaining racks must consume 6kW per rack now, otherwise they will strand capacity! Since capacity is the unit of measure for pricing in data center leasing or construction, this is a very significant change. Applications often do not drive the IT hardware to the expected level of consumption. If the 30 remaining planned 5kW racks are actually running at 4kW, that means that 2kW per rack (60kW) or 30% of the capacity is unusable in a pre-fab data hall.
As in the case of containers, Pre-Fabricated data center solutions are also slaves to their physical limitations. Although they reside within physical buildings, either the provider’s or one built by you, the data center’s themselves resemble containers in terms of their actual size. This means that the “one structure, one load group” restriction and its corresponding limitations applies.
Like containers and pre-fabricated solutions, standalone data centers are not tied to any specific geographic location like monolithic data halls. Unlike these space constrained alternatives, however, standalone facilities are actual buildings featuring raised floor areas of 10,000 square feet or more. Since these solutions are dedicated to a single customer the entirety of the raised floor is available to support all of their required applications. These facilities are also optimized to support high-density applications without the need for containment. Applications of up to 20kW per rack can be supported without the need for any special equipment. Designed to function as complete data centers, standalone facilities provide the only real alternative to monolithic modular-data halls as solutions to support mixed load groups within a single facility.
Effective use of capacity, on-going site management and the ability to effectively plan for future growth are all major elements of data center operations that are directly impacted by load group implementations. In the increasingly rare instances where homogeneous load groups are the primary data center requirement, customers have a broad array of potential solutions. Due to a variety of factors including cost, today’s customers are seeking solutions for their heterogeneous load group requirements, and these requirements can only be addressed by data centers that are not subject to the spacial limitations imposed on them by their architectures.