Data Center Design Considerations

POWER REQUIREMENTS

As recent as five years ago, planners could rely on tried and true rules of thumb to plan for an adequate amount of power for a data center build-out. However, the staggering pace of technological advancements(especially with servers, switches and storage devices) have combined to completely change the landscape for data center power consumption. 

As an example, during the five years between 1998 and 2003, data center power consumption doubled in the U.S. Even more remarkable is that this consumption doubled again in the three years between 2003 and 2006 as reported by the U.S. Department of Energy. Since 2003, the increasing acceptance and deployment of blade and 1U (single rack unit) servers are resulting in a power consumption density at the cabinet level that
drastically changes the traditional rules of thumb designers relied on for more than 15 years. 

Today, the pace of power consumption continues to rapidly increase in data centers. As a result, in late 2006, the House and Senate passed bills that tasked the Environmental Protection Agency with how to reduce the pace of power consumption in data centers.

Along with power consumption, data centers are also expected to operate without downtime. One of the ways to help data center managers achieve their company’s availability objectives is to design redundancy into the data center infrastructure.

As the demand to be operational increases, often it is necessary to duplicate components that make up the data center infrastructure, either to allow for planned maintenance or to protect availability from unplanned failures. At the upper end of availability requirements, multiple power grids are funneled into a single data center to provide assurance that there
will always be power available to run the facility.

Thanks to the efforts of the Uptime Institute and other standards resources, most active component manufacturers today build their components with dual (sometime triple) power supplies. 

Multiple power supplies allow a component to have power delivered to it from multiple power sources and pathways.

In conjunction with the increasing amount of power in data centers, it is also becoming more critical to build a solid electrical grounding grid to protect the equipment and the personnel. See ANSI/-J-STD-607-A or CENELEC EN50310 for more information.

While the traditional rule of thumb for data center power planning has typically been expressed in “watts per square foot,” availability conscious data centers are adopting “watts per cabinet” as their new rule of thumb for power planning. 

In addition, as data center managers migrate toward either Tier 3 or Tier 4 availability, they are constantly looking at ways to create redundancy in their infrastructure. This is especially true of data centers that are required to operate all day, every day. Without redundancy, data centers would have to shut down the systems to perform routine maintenance.

Even though redundancy in the infrastructure allows continuous operation of the data center, it creates significant problems at the cabinet level. Redundant power supplies require twice as many power cables and power strip outlets. Multiple communication pathways increase the number of data cables outside of the cabinet.

In addition, as the power and thermal densities increase at the cabinet level, it is increasingly important that management and monitoring systems are added to the cabinet and the data center. The systems that are available today provide complete coverage of the environmental and component operation at the cabinet level, as well as for the
entire data center. Typical systems can be configured to monitor temperature, smoke and humidity, as well as power consumption, fan status and UPS operation.

Fig 2: Basic Power System 

UPTIME INSTITUTE TIERS

Tier I (Basic) (N*):

Need single path for power and cooling distribution, no redundant components, 99.671 percent availability.

Susceptible to disruptions from both planned and unplanned activity. UPS or generator is a single-module system with many single points of failure. To perform annual preventive maintenance and repair work, the infrastructure must be completely shut down.

Tier II (Redundant Components) (N+1):

“Need plus one” single path for power and cooling distribution, redundant components, slightly less susceptible to disruptions–99.749 percent availability.

Maintenance of the critical power path and other parts of the site infrastructure require
a processing shutdown.

Tier III (Concurrently Maintainable) (N+1):

Multiple power and cooling distribution paths, but only one path active. Allows for any planned site infrastructure activity without disrupting the computer hardware
operations. Contains redundant components that are concurrently maintainable with 99.982 percent availability. Tier III sites are often designed to be upgraded to Tier IV sites when cost allows.

Tier IV (Fault Tolerant) 2(N+1):

S+S* multiple active power and cooling distribution paths, redundant components, 99.995 percent availability. The site infrastructure is able to sustain at least one worst-case unplanned failure or event with no critical load impact. S+S=System+System–two
separate UPS systems in which each system has N+1 redundancy. This requires all computer hardware to have dual power inputs. It supports high availability IT
concepts that employ CPU clustering, RAID and DASD storage methods, and redundant communications to achieve high reliability, availability and serviceability.

* Please note: the Uptime Institute defines “N” as


“Need” and “S” as “System.”



This article has been extracted from the "Data Center Infrastructure Resource Guide "by Anixter

The full free guide can be downloaded from the below link


Download the Free Guide Here