![]() |
|
![]() |
![]() |
|
EDUCAUSE Quarterly
|
![]() |
Green Desktop Computing at the University of OxfordGreen Desktop Computing at the University of Oxford
The government of the United Kingdom has set a target to reduce CO2 emissions by at least 34 percent from 1990 levels by 2020. The Carbon Reduction Commitment (CRC) will require all large public and private sector organizations across the U.K. to cut carbon emissions and report total CO2 emissions annually so that the data can be published in a league table. IT professionals clearly have a significant role to play in helping organizations perform effectively within this framework:
The project reported on in this article considers the first of these initiatives, specifically how to minimize the power consumed by desktop computers at the University of Oxford. If left on all the time, an 80 watt desktop computer and 25 watt monitor will consume 920 kWh over a year. The same equipment will consume about 198 kWh if switched off at the end of each working day. This reduction currently equates to a saving of £87 and 387 kg carbon dioxide equivalents (CO2 eq) per computer per annum. According to the Joint Information Systems Committee (JISC) publication “Sustainable IT in Further and Higher Education: SusteIT Final Report,” further and higher education organizations in the U.K. own about 1,470,000 computers, thus this simple change in power management equates to an annual saving of about £127,000,000 and 570,000,000 kg CO2 eq. Obviously there are many assumptions built into this calculation, but we think that at least half of this saving is realizable. At education sector IT events over the past two years we have asked delegates to raise their hands to indicate which best described the approach to desktop computer power management within their group:
We then asked people to raise their hands if they thought their group should adopt a policy to ensure that the campus community safely and reliably powers down idle computers. One hundred percent agreed. Initiating Green Desktop Computing at OxfordThe University of Oxford encompasses more than 100 autonomous groups (colleges, departments, libraries, museums, and so forth), most of which manage their own IT systems. The different groups have significant variations in hardware, operating systems, installed applications, and central management software (for software updates and data backup, for example). Of course, people and groups also differ in their requirements of IT services and skills. At Oxford, as in many large organizations, one approach rarely suits all. In 2006 the Environmental Change Institute asked Computing Services to develop tools and techniques for reducing the energy consumed by desktop computers at the University of Oxford. Together with the Oxford e-Research Centre and the Estates Directorate, the two groups submitted a proposal to the Joint Information Systems Committee (JISC) for funding under its Institutional Exemplars program. JISC chose to fund the project to demonstrate an approach to green desktop computing in a complex institutional environment. About the TechnologyMost computers can easily be configured to go into S3 (sleep/standby) and S4 (hibernate) power-saving modes automatically. This typically means reducing power consumption to below 5 watts. IT managers often disable this power management setting because it can:
(Lisa Hopkins and her colleagues at the University of Liverpool describe these issues in detail on the University of Liverpool’s Computing Services PowerDown web page and the British Computer Society also has a good article.) We concluded that the easiest, safest, and most reliable way to reduce IT-related energy consumption by idle computers, across all types of operating system and hardware, is to ask people to switch off their computers at the end of each working day. Turning off computers might cause problems, though, if central services need them to be on when performing backups or installing software updates; if people are unhappy waiting for their computers to boot and login scripts to run (which can take from 90 seconds to 20 minutes); or if users require remote access (from home, for instance, or when away at a conference). Actually getting computers switched off might not be easy either, if people forget or object to switching their computers off manually. In addition, it is difficult for an automatic process to know when a computer is truly idle (for example, avoid switching off a computer that has an experiment running overnight). To encourage users across the university to adopt the practice of switching computers off at the end of the working day, we developed monitoring and wake-on-LAN services (WOL).2 IT staff can provide both services to their users by installing open-source software we simply call monitoring and WOL “gateways.” Computing Services provides full instructions on how to do this, and the monitoring and WOL software is freely available under an open source license.3 MonitoringThe monitoring service uses Address Request Protocol (ARP) queries to gather and plot the number of devices with IP addresses throughout the day and night. (We use ARP because it provides higher accuracy than normal scanning protocols such as Ping.) The resulting graph shows the number of computers that are switched on throughout the day and can be used to estimate how much energy is being wasted by idle computers. People can access the monitoring service graph via a web page, allowing them to easily see how well their green desktop computing initiative is being implemented. We reason that if people can see that their effort (switching computers on and off) is being reciprocated by others, they are more likely to continue contributing to the initiative. Wake on LANThe WOL service allows computers to be switched on remotely:
Figure 1 illustrates the technical design of the WOL service. The green person icon at the left represents a user, who can log into the registration service via a web page. Users can then register and switch on their computers using the central WOL service, which relays wake up commands or “magic packets” to computers via the gateway server software. The HFS (Hierarchical File Server)4 backup service can also use the WOL service to switch computers on just before they are scheduled for backup. Figure 1. Schematic Representation of the Wake on LAN Infrastructure Five Steps to Green Office ComputingOur goal is to encourage as many groups across the university as possible to engage with green desktop computing. To achieve this, we are promoting a five-step process:
In practice this typically means:
Project ResultsThe energy manager at the Estates department sends monthly reports to building administrators detailing current energy consumption compared to the same month during the previous year. A group that moves from 200 computers permanently switched on to most computers being switched off at the end of each working day would expect to see a graph like the one in Figure 2, which shows the hypothetical reduction in electricity demand resulting from that change. Figure 2. Hypothetical Reduction in Electricity Demand of 200 Desktop Computers Switched Off Overnight Rather than Left On Energy prices in the U.K. have fluctuated significantly over the past two years (£0.08 to £0.12 per kWh) and will continue to change in the near future.5 If left on all the time, an 80 watt computer and 25 watt monitor will currently cost about £110 in energy charges and cause 494 kg CO2 eq to be released into the atmosphere. If switched off at the end of each working day, this equipment will cost about £24 in energy charges and result in 106 kg CO2 eq. A department with 200 computers left on all the time would save about £17,000 and 78,000 kg CO2 eq per annum if staff switched their computers off after work each day. Of course, many assumptions are built into this calculation, some of which are outlined on the Computing Services Sustainable Desktop Computing, Estimates, web page. Actually observing these results can be difficult because the electricity meters installed in buildings often cover energy supplied to areas with a wide range of uses and devices, such as research equipment in laboratories, office computers, and other machinery. The monitoring service allows us to report the outcome of a green desktop computing initiative by totaling the number of devices with IP addresses. Figure 3 shows real data from the computing services building over a four-week period. Each week is easy to see with the five-fingered peaks and troughs separated by weekends. The significant drop between about 280 live nodes and about 160 shows that most staff power down their computers overnight. The remaining live nodes are printers, development servers, switches, and desktop computers that cannot be switched off (for example, because their network card does not support WOL). Careful inspection of the graph shows that more computers are left on overnight on Wednesdays compared to the other days in the week. This is because computers are left on to allow them to be backed up, and the computers left on all night on Wednesdays are those that have not been configured to allow the backup service to switch them on and off either side of their scheduled 15 minute slot. Significantly, many of the colleges will see their monitoring graphs change during term time when students plug laptops into the network. We are currently unsure how students manage the power consumption of their personal equipment, but with over 90 percent of students bringing a laptop, this is likely to be a significant area of focus for a green desktop computing initiative. Figure 3. Monitoring Service Graph of Computing Services Devices with IP Addresses, July 2009 The green desktop computing project’s overall goal is to encourage as many groups as possible to engage with the five-step process. We aim to publish case studies that demonstrate how successful we have been in achieving this at the University of Oxford. Planned WorkTo take the project forward we are currently working on the following activities:
DiscussionAnyone who has been involved in implementing an IT project will know its success depends in part on the quality of the service but mostly on the willingness of people to adopt new ways of working. We have discussed the aims of our project with people from a wide range of educational and commercial organizations in the U.K. over the past two years. Table 1 provides a summary of some of the opinions expressed, with a focus on the helpful and unhelpful attitudes. Table 1. Helpful and Unhelpful Attitudes Affecting the Success of Green IT Projects
These comments highlight the fact that managers wanting to implement energy-saving initiatives will need to consider a complex set of often highly emotive attitudes. Perhaps we can learn from the struggle doctors have had with tobacco companies (and corrupt politicians) over the past few decades: it is very easy for a few people to spread doubt about scientific consensus, such as the link between smoking and cancer. Our experience suggests that the success of green projects will depend on our ability to promote better understanding and acceptance of scientific research. ConclusionsImplementing a robust approach to desktop computer power management across a complex decentralized IT infrastructure is challenging but feasible. At Oxford our approach has been to encourage building administrators and IT staff to work through the five step approach described above. Where automatic power management is not feasible, achieving office-computing energy reductions will depend on many people “doing their bit,” which can be seen as both a problem and opportunity: manually powering down a computer will be an annoyance for some people, but for others it will provide a tangible way to demonstrate their understanding of the issues associated with IT and the environment. Initial results are encouraging but we still have work to do to achieve the levels of saving we want to achieve. If you would like to learn more about our work, please send e-mail to greenit@oucs.ox.ac.uk. AcknowledgmentsWe would like to thank:
Endnotes
© Howard Noble, Daniel Curtis, and Kang Tang. The text of this article is licensed under the Creative Commons Attribution 3.0 license. |
![]() |
||||||||||||||||||||||||||
| Unless otherwise noted, EDUCAUSE holds the copyright on all materials published by the association, whether in print or electronic form. In certain cases the work remains the intellectual property of the individual author(s) (see Special Circumstances). Content from conference speeches, presentations, blogs, wikis and feeds reflect the opinions of the author, and not necessarily those of EDUCAUSE or its members. | |||||||||||||||||||||||||||||