Tucked away on pages 54 and 55 of the EPA's recently released report on server and datacenter efficiency is a rather impressive and comprehensive list of strategies for reducing energy costs in your datacenter facilities.
While many action items are certainly not as simple as, say, adjusting the thermostat a bit or plugging holes in the datacenter floor, it would behoove any datacenter operator to run down the list and at least do some research on those items that seem feasible or at least worth contemplating.
One of the more intriguing proposals, to me, is at the top of the list: "Design software to avoid excess code and inefficiencies." The topic of software in the context of energy efficiency is one that I hope to explore in the near future. I can't help put wonder if it contributes to the increasing interest that we're seeing in Linux in the datacenter.
Following is the list of "Potential Energy-Efficiency Improvement Opportunities for Servers and Data Centers" as outlined by the EPA. (I've sprinkled in some links to InfoWorld coverage for some of the topics.)
Computing software
1. Design software to avoid excess code and inefficiencies (treat CPU cycles as a finite resource)
2. Provide developer tools to help improve efficiency of software
3. Enable shifting of computational load among systems for maximizing energy efficiency
4. Upgrade applications no longer supported on latest technology and/or operating systems, allowing removal of legacy servers
5. Implement virtualization to allow consolidation of server and storage hardware
IT hardware (computing, storage and network)
Operational Improvements
1. Turn off (ideally remove) dead, obsolete, or excess equipment
2. Turn off or power-manage equipment that won't be used for extended periods of time (e.g., development systems not in active use, systems for future expected increases in activity, etc.)
3. Enable power-management features on existing equipment (e.g., frequency voltage scaling)
4. Maximize utilization of storage capacity through shared data storage, data compression, and data de-duplication
Design improvements
1. Accept high-efficiency power supplies over full operating range (including DC-DC conversions) or directly accept moderate DC voltage
2. Digitally control power supplies to better match output to load
3. Use high-efficiency variable speed fans (within IT equipment)
4. Reduce energy use at lower utilizations (whether the resource is processing capacity, memory, communications, or etc.). Applies to individual systems and to clusters.
5. Improve microprocessors to lower leakage current, increase system integration, etc.
6. Use storage virtualization and massive array of idle disks (MAID) technologies to allow storage power management
7. Use centralized servers (large systems) to improve sharing of computer resources
8. Improve hardware support for virtualization
9. Use built-in power monitoring
Electrical Systems
1. Use high-efficiency power distribution (i.e., higher-voltage AC or moderate-voltage DC (50-600 VDC))
2. Use premium-efficiency motors in fans and pumps
3. Use high-efficiency UPS units over full range of load
4. Use rotary-based UPS units
5. Right size power distribution and conversion to optimize efficiency
6. Use on-site generation with grid as back-up Heat Removal
1. Improve airflow management (i.e., use hot/cold aisle configuration or penetration sealing)
2. Adjust environmental conditions (temperature and humidity set points) to allow wider range while still meeting manufacturer specifications
3. Optimize data center airflow configuration using visualization tools (computational fluid dynamics modeling or infrared tomography)
4. Use high-efficiency variable-speed air-handler fans and chilled water pumps
5. Use variable-speed chillers
6. Use variable-speed, primary-only chilled water pumping
7. Use high-efficiency chiller and chilled water supply motors
8. Use high-efficiency CRAC units
9. Use air-side economizers (outdoor air) when outdoor conditions permit (preferably in mild climate locations)
10. Use water-side economizers (cooling tower) when outdoor conditions permit (preferably in mild climate locations)
11. Commission infrastructure systems to ensure set points are at proper values, sensors are in calibration, airflow is within design tolerances, etc.
12. Rebalance air-handler system after significant IT reconfiguration
13. Size systems and configure redundancy to maximize efficiency e.g., use redundant air-handler capacity in normal
14. Increase chilled water supply and return temperature difference to reduce chilled water flow
15. Optimize chilled water plant (cooling tower) Reuse waste heat for space heating
16. Use direct liquid cooling (water or other dielectric liquid) with currently available technology (i.e., in-rack or in-row cooling) and emerging technology (i.e., in-chassis or chip-level) Controls and Management
1. Use system management hardware/software that enables powering down (to sleep and/or off) parts of server clusters during times of low utilization
2. Dispatch non-time-sensitive computational operations to reduce peak computing load and allow reduction of total
3. Provide for standard reporting of power use, platform temperature, and processor utilizations to assist operators in understanding and managing energy use in their data centers
4. Use shared computing models, such as grid computing
5. Optimize cooling controls to dynamically match the cooling supply to the IT heat load
6. Dynamically optimize the assignment of work across the data center to ensure maximum efficiency
7. Monitor power in real time
Distributed Generation
1. Use combined heat and power
2. Use renewable energy (e.g., photovoltaic panels)
3. Use fuel cells