HP BL260c HP BladeSystem c-Class Site Planning Guide - Page 30

Power consumption Power consumption can be divided into two broad categories: • Marked electrical amperage, which is listed on the required safety and regulatory labels, generally represents the maximum current draw that the marked device could achieve. Safety and regulatory labels on computer equipment list the ratings for maximum power consumption conditions. • Typical consumption, which is measured under normal circumstances, should be used only in customer calculations with UPS and air conditioning sizing where remaining capacities are needed. When determining the necessary electrical infrastructure required, consider several factors. The first and most important area is the local electrical and regulatory codes. The installation must meet or exceed these codes. The local authority has jurisdiction and makes the final decision as to whether an installation conforms to the relevant codes. In North America the relevant code is National Electrical Code 2005 Article 210 and Article 645, which states the continuous measured load should not exceed 80 percent of the rated circuit maximum, where continuous load is defined as 3 hours or more. This method leaves some flexibility to the infrastructure designer and includes several options for determining power requirements, which are discussed in the following bullets. NOTE: The following scenarios use North America circuit sizes and voltages. • Size the electrical infrastructure according to the label ratings, even though this could mean the electrical infrastructure might be oversized. Typically the power supply is sized to provide the maximum power consumption throughout the life of system. Additionally the power supply might be used in multiple systems, and therefore must be sized to the maximum configuration of the largest system into which it can be installed. Sizing the electrical infrastructure to the power supply label rating protects against ever overloading the electrical infrastructure, but typically results in overprovisioning. As an example, due to the long lifecycle of the c7000 enclosure (5 years) the power supply is rated at 2400W output, 2780VA input. When running in N+N redundant mode, according to the power supply specifications, the system is rated at a maximum input power of 8340VA (2780VA x 3). To deliver this much redundant power to an enclosure requires either two 50A single-phase circuits (8320VA each) or two 30A three-phase circuits (8640VA each). Some extreme configurations can simultaneously stress all 16 server blades to maximum load, which would use the full capacity of the power supplies. An order of magnitude calculation using the power supply ratings requires approximately 2MW (8kW x 3) to be delivered to the racks. Full redundancy would require 4MW. The cost of the UPS equipment, transformers, switchgear, generators, and other infrastructure to provide 4MW is significant. The Uptime Institute (http://www.uptimeinstitute.org/) estimates that each watt of redundant power costs between $23 and $25 depending on the level of redundancy required. For typical environments, power usage would be between 750kW and 1.25MW, or 40% to 60% of the available capacity. Actual power usage is significantly lower because the vast majority of configurations running typical customer applications do not generally exceed 5kVA peak power consumption. The average power consumption is usually lower, typically between 3kVA and 4kVA. Sizing using nameplate values typically results in significant amounts of wasted power capacity. Power requirements and considerations 30