Why the End of Moore’s Law Will Drive Energy Efficient Data Centers
The March 12 edition of The Economist Magazine declared that after 40 years “the End of Moore’s Law is in sight”. Moore’s Law—actually more like a rule of thumb—was first expressed in a 1965 paper by Gordon Moore, an Intel co-founder. In this paper Moore noted that the number of components placed on an integrated circuit was doubling every year. In the 70’s this rate of doubling occurred every two years at no rise in per-unit cost and this has continued for the past 45 years.
This continuous improvement arc in computing has created remarkable advancements in the ability to store and process data:
- In 2000, 800,000 Petabytes (PB) of data were stored in the world. By 2020, 35 zettabytes (ZB) are expected to be stored—- a 44,000 + order of magnitude increase.
- 90% of the world’s data was created in the past two years
- Smartphones have more computing power than supercomputers developed in the 1980’s
While the doubling of computing capability every two years is showing signs of slowing, Linley Gwennap, President of Linley Group, a technology consultant firm, argues that “from an economic standpoint, Moore’s Law is over”. Put differently, the chip-making industry is no longer able to keep costs flat when doubling computing power. This places pressure on Data Center operators to lower costs in other ways—like lowering energy usage.
The Rise of Data Centers and the Cloud
Individuals and companies have shifted their centers of high-powered computing from personal computers to the computing “cloud”. In reality, the cloud is an ever-growing legion of data centers, operating within firms for dedicated internal use, or, increasingly, owned by data center operating specialists as a service to large firms and individuals.
According to Cisco Systems, in 2012 the cloud-generated proportion of data center traffic was 1.2 zettabytes and comprised 46% of all data center traffic. By 2017, Cisco estimates that the cloud will be generating 5.3 zettabytes and make up 69% of all Data Center traffic.
Data Centers and Energy Use
While data centers minimize up-front user costs, they are large consumers of electricity. Data Center buildings are 10-100 times more energy intensive than office buildings. Global Data Center electricity consumption is anticipated to grow from 2% of total electricity consumption at present, to 3-4% by 2020.
The management of the building climate in which a data center operates is critical to energy cost-effectiveness and computing reliability. Too much climate control and data centers become refrigeration units. Too little climate control and risks such as diminished speed and component burnout are introduced.
Furthermore, the strategic geographical location of new data centers will become more important. Data Center Operators will seek out locations in climates where they can draw from a cooler ambient environment to help condition inside space. A limiting factor to “cool climate” concentration of data centers is that data centers must connect with the users they serve, and in most cases these users require maximum speed; speed which can be compromised over a distance of miles, much less oceans.
As with all Energy Efficiency projects, establishing reliable benchmarks and baselines for Data Center energy efficiency projects is very important. Power Usage Efficiency (PUE) is an energy metric that provides an indication of the efficiency of data center infrastructure systems and is the most globally recognized performance metric for the energy efficiency of data centers infrastructure support systems. PUE is the ratio of total amount of energy used by a computer data center facility to the energy delivered to computing equipment (IT Load). The average PUE for Data Centers has fallen gradually from 1.89 in 2012 to 1.70 in 2014. While servers and other IT equipment are the main power users (IT Load), the amount of energy required to cool and deliver the air and provide lighting to the IT equipment amounts to about 50% of a typical data center’s energy consumption.
Specific Technologies to Improve Data Center Energy Efficiency
Through its experience in working on Data Center energy efficiency projects, DNV GL has identified a number of best practice technology approaches to improving Data Center PUE:
- Airflow management
- Airside management containment
- Airside HVAC, in-row cooling
- Rational Data Center thermostat set-points
- Use of “free” ambient cooling
- Strategic server rack configuration
- Optimized power distribution within a Data Center
- Data Center Infrastructure Management (DCIM)
- High efficiency multi-core processing servers
- Server virtualization
- High Efficiency Power Distribution Units (PDU’s) and power supplies
That computing activity will continue to grow dramatically in the coming years is virtually certain, even against a backdrop of upward cost pressure in chip manufacturing. The Data Center market will grow significantly and become a larger portion of an electricity service provider’s portfolio. Customer engagement initiatives must uniquely address the Data Center market segment or risk losing this load. A key element of an electricity service provider’s strategy for addressing Data Centers’s needs will be an energy efficiency offering. This offering should engage data center managers through technical expertise and winning technology approaches; knowledge of what works and what does not; local knowledge (e.g. climate, weather patterns, load curves, etc.); and financial incentives to encourage deeper and long-lived energy efficiency measure adoption.
DNV GL Energy’s Data Center Energy Efficiency Services
DNV GL has successfully implemented Data Center Energy Efficiency projects on behalf of utility clients. Our team has significant experience on both new construction and retrofit data center energy efficiency projects. We work through the entire cycle of project identification, justification, and valuation and follow this up with detailed engineering analysis and post installation review, assuring that the promise of a Data Center Energy Efficiency project is fully realized. Our team is also available to work directly with Enterprise Wide Data Center Operators to assist them in Company wide energy efficiency initiatives.
For more information, please visit our Data Centers Knowledge Hub or contact John.Greco@dnvgl.com.