Purchasing More Energy-Efficient Servers, UPSs, and PDUs
New servers use more energy-efficient technology such as:
- More efficient power supplies, better DC voltage regulators, processors that consume less power, and cooling fans that are more energy-efficient.
- CPU speed throttling and power saving technology that reduces the processor's clock rate and voltage when the processor is idle. The aim is to minimize overall power consumption and lower heat generation, allowing for slower (and quieter) cooling fan operation.
- The latest operating systems, which take advantage of hardware features to deliver optimal power efficiency for any given workload running on the server. By default, Windows Server 2008 R2, for example, continuously alters the power states of the processors in the system in response to the utilization level of the workload. This ensures that processor power usage maps to the needs of the workload, with minimal impact on performance. R2 achieves additional power savings by combining processor power state control with a feature that consolidates work onto a smaller number of processor cores when workload utilization is low. This feature, referred to as Core Parking, places unused processor cores into a deep sleep state, effectively scaling the number of processor cores in active use. Other features such as Timer Coalescing and Intelligent Timer Tick Distribution (or Tick Skipping), extend the time that processor cores stay in deep sleep states by avoiding waking cores unnecessarily. R2 also includes interfaces for remotely accessing power meter and power budget information and for modifying Windows Power Policy across groups of servers.
The ENERGY STAR Server specification, effective May 15, 2009, requires servers that display the ENERGY STAR label to have:
- More efficient power supplies,
- Improved power quality, which provides building-wide efficiency benefits,
- Capabilities to measure real-time power use, processor utilization, and air temperature,
- Advanced power management features, and
- Data sheet for purchasers that standardizes key information on energy performance, features and other capabilities.
Uninterruptible Power Supplies (UPSs)
- UPSs maintain power to electrical loads in the event of a utility power supply disruption lasting from just seconds to tens of minutes. In addition, UPSs condition the power reaching the load under normal operation to prevent undesired features of the power source (outages, sags, surges, bad harmonics, etc.) from adversely affecting the performance of servers and other equipment. UPSs typically use batteries as an emergency power source and provide power to servers until emergency generators come on line.
- The primary energy waste in a UPS is the switching losses in the inverter and transformers. To mitigate switching losses, a power management system precisely controls every pulse of the switching cycle, optimizing the inverter's switching for specific load types and load levels. The resulting switching patterns suffer the least possible losses and surpass the efficiency of older-style systems with fixed switching patterns. In addition, UPSs can offer a 2% to 3% overall efficiency advantage over a generic lower efficiency transformer due to their high-efficiency transformers.
- New efficient UPSs generally range from 92% to 95% efficient.11 The Eaton Energy Saver System claims to reach 99% efficiency across a wide range of loads.12
- An ENERGY STAR qualified UPS can cut energy losses by 30-55%. A 1000 kVA UPS used in a large data center could save $18,000 annually. See the list of ENERGY STAR-qualified UPSs.
New UPS Systems Save Energy in Exelon Data Center
Exelon has made a number of improvements to its Joliet, IL Data Center over the years, but arguably no single improvement was more effective at boosting overall data center efficiency than upgrading the UPS and power distribution systems. About five years ago Joliet, a Tier II data center, operated two 1000 kVA Piller UPSs in parallel. These large UPS systems were inefficient because they ran at such a low load capacity. The power distribution board at that time was also sub-optimal as it represented a single point of failure.
Today the data center is served by four smaller (750 kVA) UPS modules configured as two x 2N systems. This new UPS configuration offers full redundancy (meaning that if one UPS system fails, a redundant UPS system can supply all of the power for critical loads) -- and the ability to add capacity as needs grow. The efficiencies of the individual UPS units are above 90% due to high load capacity. Power is distributed by independent paths to pairs of PDUs. Each PDU pair receives power from at least two UPS modules. Additionally, monitoring equipment is installed in all PDU and RPP distribution panels to measure system efficiency and performance.
"Replacing the old UPS systems was probably our biggest challenge on the path towards a more efficient facility," recalls Dom Lovino, Exelon Corp's IT Manger of Data Center Operations. "But it also provided one of the biggest payoffs in terms of efficiency." These changes and several other efficiency upgrades (new CRACs with variable speed drives and dual coils, airflow management improvements, and an increase in the allowable temperature and humidity range) brought the data center's PUE down from 1.96 to 1.146. Although computing capacity has increased at least 5-fold during the past five years, overall energy consumption at the Joliet Data Center has remained virtually flat. Exelon tapped The Green Grid, DataCenter Dynamics, Uptime Institute, and nlyte Software for assistance in identifying and implementing energy efficiency measures at the Joliet Data Center."
Power Distribution Units (PDUs)
PDUs deliver conditioned power from the UPS to servers, networking equipment and other electronic devices. PDUs have two meanings in data centers:
- PDUs can simply be well-constructed power strips suitable for data center use. In addition to basic PDUs, switched rack PDUs provide load metering combined with controlled on/off switching of individual outlets. Providing real-time power data, metered PDUs provide a way to understand data center loads and the corresponding potential for hot spots.
- Floor-mount PDUs traditionally contain isolation transformers and feed power downstream to rack-based PDUs. Step-down transformers reduce high voltage and current to more common and useful levels; for example, from 240V 30A single phase to multiple 120V 15A or 120V 20A plugs. They can be integrated with monitoring software to trend power consumption and plan for future power needs. Like UPSs, PDUs can use high-efficiency transformers that are 2% to 3% more efficient overall compared to a generic lower-efficiency transformer.
Target, in one of their two ENERGY STAR certified data centers, powered down two unloaded 300 kVA PDUs to save 261,000 kWh annually.
Savings and Costs
- A recent set of tests conducted jointly by US EPA, HP, and Microsoft demonstrate that replacing an older server with a new ENERGY STAR-qualified model will save energy and deliver more processing power in the bargain. In some of these tests, the new ENERGY STAR server consumed 54% less power than older model servers. Servers that earn the ENERGY STAR will, on average, be about 30% more energy-efficient than standard servers. Learn more (PDF, 2.67MB).
- At the average US commercial rate for electricity of 10 cents per kilowatt hour (kWh), the energy savings from a single ENERGY STAR-qualified server could range from $60 (at 50% utilization) to $120 (at idle) annually, or $240–$480 over the useful life of a server (4 years).
- In addition to using less energy themselves, ENERGY STAR-qualified servers substantially reduce cooling loads in data centers. A general rule of thumb suggests that one watt saved by a server has the added benefit of saving one to two watts of cooling power. This yields a total savings of between $480 and $1,440 over the useful lifetime of a server. It's important to note that these power savings come with a substantial increase in performance—at 50% utilization, for example, the newer, more energy-efficient server handles over three times the workload, thereby reducing the number of systems needed to support the same load.
- New servers,
- Labor costs for removing old servers and deploying new servers, and
- End-of-life disposal costs for systems that are not repurposed.
- DOE estimates that a 15,000-square-foot data center operating at 100W/square foot would save $90,000 by increasing UPS efficiency from 90% to 95%.13
- New UPS,
- Downtime to reconfigure electrical load,
- Labor costs for removing old UPSs and deploying new UPSs, and
- End-of-life disposal costs for systems that are not repurposed.
- In some cases it will be necessary to manually enable processor power saving features such as:
- Intel's version: Enhanced Intel SpeedStep Technology (EIST) or Demand Based Switching (DBS), and
- AMD's version: Cool'n'Quiet.
- When installing new high-density servers, consider using in-rack and in-row cooling, if possible, including direct liquid cooling.
- Proper preparation can substantially mitigate the risk of downtime when migrating applications and files from an existing server to a new system.
- UPS systems rarely run at 100% load. UPSs in N+1 or 2(N+1) redundant configurations hardly ever exceed 30% load per module.14 The critical point for evaluating is partial load efficiency, which is 70% for most Tier I-III systems and 30% for most Tier IV systems. Unfortunately, the efficiency of a UPS is usually quoted at 100% rated load and at other favorable times, such when the system's batteries are fully charged, its UPS input voltage is nominal, and when optional input transformers and filters are disconnected or not installed. A more accurate evaluation would be the UPS's efficiency at roughly 30% to 70% load, which is the average load for most medium- to large-scale data centers.
13 Best Practices Guide for Energy-Efficient Data Center Design, NREL, February 2010.
14 N+1 redundancy is a form of resilience that ensures system availability in the event of component failure. Components (N) have at least one independent backup component (+1). N+2 represents the addition of still one more component to serve the load and permits any piece of equipment to be removed from load for scheduled preventive maintenance while retaining the ability to tolerate the failure of a piece of equipment with no loss of service. 2N represents the arrangement and number of systems necessary to serve the load of the computer center (cooling, power, etc.) while also providing the ability to remove a system from service or repair with no loss of service. 2 (N+1) represents dual systems, each system having redundant components.