Generally speaking, newer servers have more energy-efficient features than older servers. These features might include more efficient power supplies, better DC voltage regulators, processors that consume less power, cooling fans that are more energy-efficient, and built-in power management features.
Power management refers to the dynamic management of a server component’s performance in order to reduce power consumption for any given workload. Power management features save energy, especially for applications that do not run continuously or are accessed infrequently. For example, server processors can reduce power consumption by automatically adjusting the speed of the CPU during times of low utilization, or by automatically disabling CPU cores temporarily (when they are idle).
Check for power management options that come with your server hardware and your server operating system, and enable these features when possible:
Also known as “dynamic voltage and frequency scaling,” CPU throttling automatically reduces the processor's frequency (or clock speed) and voltage when idle. The aim is to minimize overall power consumption and reduce heat generation, allowing for slower cooling fan operation.
This feature dynamically disables CPU cores in an effort to conserve power when idle. Disabled cores are re-enabled as the CPU load increases once again. 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.
Power profiles are pre-defined groups of power management settings. You simply choose a power profile and the server configures the corresponding power management options (which may include some of the features listed above) appropriately. Power profile names and underlying configurations vary from brand to brand, but they typically provide you with three basic options.
- Balanced. This mode enables the power management options that have a negligible impact on system performance.
- Maximum Performance. This power profile delivers maximum performance.
- Maximum Energy Efficiency. This setting minimizes system power usage by enabling power reduction mechanisms that may also affect performance.
Power Metering and Budgeting
Some server software platforms support power metering and “budgeting” – setting power limits, or caps, for individual servers. Power use is monitored via a power meter (a hardware component inside the server) that reports server power consumption in watts. This information is typically provided as part of a power supply or by using a baseboard management controller (BMC).1
Power budgeting information is used to determine the power limit, or budget, that can be consumed by a server. Depending on the hardware platform, this information might also allow you to cap the system's power budget. Dell’s OpenManage Power Center is one example of a software solution designed to help you measure and manage server power consumption in your data center – see Figure 1, below. Microsoft’s website provides an overview of power metering and budgeting in Windows (including Windows Server) operating systems.
Savings and Costs
Energy savings attributable to power management features will vary greatly with hardware, software, feature characteristics, and workload. However, some studies have demonstrated the savings potential of some of these features:
- One recent study found that CPU throttling (dynamic voltage and frequency scaling) could reduce CPU energy consumption by up to 55% during computations2
- A straightforward examination of power consumption associated with each of three power profile settings available on Dell PowerEdge servers (power saver plan, balanced power plan, and high performance plan) found no difference power consumption when the server was idle. When the processor was fully loaded (“pegged”), the power saver plan used 18% less electricity than the high performance plan3
- A power managed server produces less heat (which reduces cooling costs) and requires less power (which reduces power infrastructure losses). In fact, as a result of these indirect energy benefits, saving one watt-hour of electricity at the server level typically results in an additional 1.9 watt-hours of electricity savings at the facility-level!4
Tips and Considerations
Servers are not always configured by default for optimal energy efficiency. Check with your hardware manufacturer and your server operating system provider to see if you are taking full advantage of hardware and software features that will reduce power consumption and save energy. Some brand-specific resources are provided below.
- Cisco’s Server Management Guide includes a chapter called Power Management for Blades and Power Capping.
- Dell offers a Power Efficiency “How To” for the Dell PowerEdge Server Portfolio (PDF, 1.5 MB). It explains how to configure Dell PowerEdge servers for optimal energy efficiency.
- HP’s technology brief, Power efficiency and power management in HP ProLiant servers (PDF, 559 KB), explains built-in power efficiencies, configurable power management options, and power monitoring & control.
- IBM provides instructions for setting “idle power saver” using its firmware, Advanced System Management Interface (ASMI). Enabling the idle power saver function forces the system to use less power when certain thresholds are met.
- If you would like to utilize server power management features in Windows Server but have concerns about performance tradeoffs, Microsoft offers Performance Tuning Guidelines. These guidelines (available for Windows Server 2008, 2008 R2, 2012, and 2012 R2) will help you tune the server settings to obtain incremental performance or energy efficiency gains.
- Red Hat offers a power management guide for Red Hat Enterprise Linux 6.
1 A baseboard management controller (BMC) is a specialized service processor that monitors the physical state of a computer, network server or other hardware device using sensors and communicating with the system administrator through an independent connection.
2 A. Paulin Florence, V. Shanthi, and C. B. Sunil Simon. Energy Conservation Using Dynamic Voltage Frequency Scaling for Computational Cloud. The Scientific World Journal Volume 2016 (2016), Article ID 9328070.
3 Windows Power Plans and CPU Performance, by Glenn Berry. See https://sqlserverperformance.wordpress.com/2016/06/08/sql-server-diagnostic-information-queries-for-june-2016/.
4 New Strategies for Cutting Data Center Energy Cost and Boosting Capacity, Emerson Network Power presentation, 2012, p.8. https://www.vertivco.com/globalassets/documents/blog/is03947_2012_energylogic_fin_62359_0.pdf (PDF, 1.4 MB).