In August 2007, the U.S. Environmental Protection Agency released a "Report to Congress on Server and Data Center Energy Efficiency." The key finding and recommendation of the report: Finding: The energy used by nation's servers and data centers is significant: 61 billion kWh in 2006 (1.5% of the total United States electricity consumption). To put it into perspective, that is more than all televisions in the US or a similar amount used by 5.8 million average U.S. households. If the status quo continues, by 2011, data centers will consume 100 billion kWh. Recommendation: The U.S. must improve server and data center efficiencies or face increased emissions, including greenhouse gases, from electricity generation.

Efficiency ScenarioElectricity Consumption Savings (Billion kWh) C02 Emissions Saving (million metric tons) Improved2315 Best Practice6038 State of the Art7447

What drives the increasing energy consumption? 1. Demand for more servers, more data storage and the supporting power, cooling, and network infrastructure.

What drives the increasing energy consumption? 2. Higher density servers with more processing power due to multi-core processors and smaller hard drives. (compare the difference in the servers above that are only two years apart in age)

In September 2007, Gartner released its Data Center report "The Calm Before the Storm": Key Findings: 1. By 2011, more than 70% of U.S. enterprise data centers will face tangible disruptions related to floor space and energy consumption. 2. During the next 5 years, most U.S. enterprise data centers will spend as much on energy (power and cooling) as they will on hardware infrastructure. 3. Upward-spiraling infrastructure demands and increasing energy costs mean that the energy proportion of IT costs could double by Key Recommendations: 1. Start monitoring the energy consumption of their data centers and modeling floor capacity to quickly make strategic decisions about the nature and importance of the data centers. 2. Where data center refurbishment or new builds are required, user must ensure that they and their engineering partners use the evolving modeling and monitoring tools. 3. Technologies such as virtualization and dynamic workload management must be implemented quickly to improve the use of assets and defer the procurement of new hardware as a mechanism to optimize floor space and energy costs.

What are we doing to meet the demand for more servers and data storage yet improve our efficiency?

1. Transitioning from direct-attached hard drives to a Storage Area Network (SAN) Direct Attached StorageStorage Area Network (SAN) Hard drives are physically connected and cannot be shared with any other server. Hard drives are connected to each server through the network and can be shared. Must correctly purchase and forecast capacity needs.Capacity can be purchased as needed. Spare capacity is isolated and wasted, resulting in underutilized hard drives consuming power. Spare capacity is dynamically allocated as servers need it, resulting in no wasted space.

2. Transitioning from traditional server hardware to the latest design technology, “blades”. Traditional Server HardwareHP c7000 Blade System Max capacity of 42U rack: 42 1U serversMax capacity of 42U rack: 64 servers in 4 10U c7000’s Poor scaling of power supplies. 42 servers x 2 power supplies each = 84 Power cannot be shared with other servers. Better scaling of power supplies. 4 c7000’s x 6 power supplies each = 24 Power is dynamically shared with all servers in chassis. More heat dissipation from inefficient power supplies. Poor air flow through rack. More cooling and air flow movement is needed. Less heat dissipation from inefficient power supplies. Better air flow engineered into chassis design. Less cooling and air flow movement is needed.

3. Virtualization – Using one physical server to create multiple, isolated server environments each with their own hardware, operating system and application emulations. One Traditional Physical Server One Host Physical Server Multiple Host Physical Servers 1 Physical Server1-10 Virtual Servers, 1 Physical100’s Virtual Servers scaling at 10:1 Application SoftwareVirtualized ApplicationVirtualized Applications Operating SystemVirtualized OS Physical HardwareVirtualized Hardware

4. Consolidation – Replace traditional server hardware with a consolidated, highly-efficient model. Virtual Servers on Blades utilizing SAN Storage 64 physical server blades. Add up to 640 virtual servers, no additional physical space requirements. 24 power supplies. No additional power requirements. Virtual Servers connect to the SAN and request only as much storage as need. No wasted or idle hard drives.

Metrics Since June 2008WattsBTUs 24 Physical servers and storage retired. (32,316)(87,103) Physical blades and SAN added to replace above hardware and support new virtuals. 19,41540,058 Physical power consumption net(12,901)(47,045) 80 virtual servers added to date with an estimated savings of 825W/ 3500BTU ea. compared to physical (66,000)(280,000) Physical + virtual consumption net(78,901)(327,045) Annual Savings of 75,881

Out-of-the-box comparison of energy consumption using the same hardware

Full SizeSmall Form FactorUltra Slim 365W240W135W 17H x 7W x 18D4H x 13W x 15D2.5H x 9W x 10D

$ V $ V

Modern Languages Lab = 30 computers Maximum use YTD for Spring ‘09 = 15 computers

Proper E-Waste Disposal What happens to all the old servers, computers, laptops, monitors, printers, keyboards, and mice? Employee resale program for working computers and laptops. Some items are used for loaners or spare parts. Remainder is palletized and plastic-wrapped for transportation.

Pallets are picked up by local vendors who disassemble, recycle and ensure no materials end up in landfills. Arc International Fortune 500 corporation with facilities in six states, processed a combined 60.2 million lbs of e-waste in 2007.

What’s next? Potential Future Initiatives 1. Enterprise Power Management for Computers- The use of enterprise software to provide better flexibility and management of the power settings for computers. In concept, this software would allow you to schedule and selectively place computers into a power savings mode when idle. For example, if we identified that all computers on the 2nd floor of building X were not needed between 10AM and 2PM, but needed to used at 3pm. We could build a schedule to allow them to enter a low power state at 10AM, but return to normal mode at 3pm. This software is still emerging area and may need further maturation before implementation. 2. Virtual Desktop Infrastructure (VDI) - This solution takes the virtualization concept and applies it to the desktop. Rather than deploying desktop computers with 240W power supplies, we could deploy thin client computers with 50W power supplies that need only enough power to connect to a virtual desktop in the datacenter. In concept, this would allow us to implement the same efficient scaling for desktops as we are doing for servers. This is still an emerging concept and ay not be an adequate replacement for all situations. 3. Printer Consolidation – Reducing the large number of personal printers on campus and replacing them with multi-function workgroup printers to reduce power usage, consumables and the volume of cartridge disposal. Vendors can perform on-site assessments, complete with floor plan mappings of printer locations and replacement recommendations for multi-function, multi-user devices. However, further analysis should be done to before any implementation to verify the savings.