- Pierre Delforge, Senior Engineer at Natural Resources Defense Council (www.nrdc.org)
As a growing number of businesses consider moving computer applications from on-premise server rooms to the “cloud,” a groundbreaking analysis released today by the Natural Resources Defense Council, in partnership with WSP Environment & Energy, reveals that not all of Internet-based computing platforms are created equal: some clouds are “greener” than others.
While mammoth “cloud” data centers are in the media spotlight for their huge energy appetites, smaller server rooms and closets are responsible for over half of U.S. server energy consumption – but 50% of that is wasted due to lack of knowledge or incentives to save, costing U.S. businesses up to $2 billion in unnecessary electricity bills. Small server rooms and closets can be found both in small- and medium-organizations (SMOs), such as services and manufacturing sector companies, hospitals, government agencies and educational institutions, as well as large enterprises where they are often spread throughout headquarters and regional branch locations.
NRDC found that as much as 30% of a typical small office-based organization’s electricity bill may be due to powering and cooling servers running around-the-clock even when performing little or no work.
Cloud computing is often touted as the most eco-friendly choice. Until today’s report, however, there had been no independent analysis validating whether these computing services delivered over the internet are the most sustainable option for SMOs seeking to improve the energy and carbon efficiency of their Information Technology operations.
NRDC and WSP’s analysis found that:
· Cloud data centers using energy-efficiency best practices and powered by renewable energy or efficient natural gas power plants can have dramatically lower carbon footprints, by as much as 97%, than typical server rooms in small- and medium-sized organizations.
· But “brown” clouds that do not optimize energy efficiency and use electricity from coal-fired power plants, can have a larger energy and carbon footprint, by up to a factor of two, than on-premise server rooms using effective methods to improve energy efficiency and sustainability.
The report, Is Cloud Computing Always Greener? Finding the Most Energy and Carbon Efficient Information Technology Solutions for Small- and Medium-Sized Organizations, demonstrates that the carbon footprint of business computing is highly dependent on a number of important variables, including the type of electricitypowering the data center, the amount of server processing capacity being effectively utilized, and the efficiency of the facility’s cooling.
We found that the factors with the biggest impact on the environmental footprint of server computing are, in order:
· Server utilization factor: The U.S. Environmental Protection Agency estimates typical servers operate on average in a range of 5 percent to 15 percent capacity while drawing 60 percent to 90 percent of their maximum power. Running more than one application on a server through a process known as “virtualization,” or having more than one customer share a server as in the case of cloud computing, can increase server utilization to 50 percent or higher. Unfortunately, most servers run a single application.
· Electricity carbon emissions factor: This is the carbon footprint of the type of electricity used to power the data center (i.e., generating electricity from renewable energy produces far lower greenhouse gas emissions than coal-fired power plants).
· Power Usage Effectiveness (PUE): This measures the efficiency of the facility housing the servers, including cooling, power distribution, and lighting.
· Hardware efficiency: Upgrading to newer models can substantially reduce energy consumption by server, data storage and networking equipment. Computing efficiency of servers is doubling every one and a half years.
Our study compared the above variables under five of the most common business computing scenarios: on-premise servers with no virtualization (servers running a single application); colocation (
hosted in external facilities shared with other companies) with no
virtualization; on-premise with virtualization; private
consolidated and applications accessed across a company’s intranet); and public clouds. As
illustrated below, a public cloud nearly always produces far less carbon
emissions and on-premise server rooms with machines operating a single
application produce the most.
Fortunately, as our analysis shows, small- and medium-sized organizations looking to improve the sustainability of their operations have multiple computing options and should review all the variables that contribute to the carbon impacts of these options. When considering the cloud, SMOs should request full disclosure of the carbon-efficiency of the services. If keeping their computing platform on-premise, businesses should make sure their servers – and the rooms housing them – work at optimum efficiency.