This study finally culminated in the submission to the JISC under the project name Low Carbon ICT (May 2007). <> == Summary == * We cannot find evidence to support the the theory that turning a PC on and off once a day will pose an increased risk of causing a hardware failure. * It is technically possible to manage all scenarios (e.g. during the day, night time, for computers that are currently on or off) for installing software remotely (to install security patches) but OUCS does not have an integrated package to allow colleges and departments to take advantage of this easily. * All monitors at least (not PCs) can be switched off every night, this web site shows power consumption of PCs monitors: http://www.energystar.gov/index.cfm?c=power_mgt.pr_pm_step1 * While users can do their bit management needs to establish best practice for instance in buying energy-efficient machines. * Big businesses like [[http://www.cerncourier.com/main/article/45/9/30|Google]] are looking at their power usage as they are consuming vast amounts of electricity. * It is likely that we need to look at different computer architectures to tackle the problem effectively. (See [[http://www.oss-watch.ac.uk/events/2006-04-10-12/presentations/simontindall.pdf|page 14]] onwards in this presentation by Sun Microsystems at a recent [[http://www.oss-watch.ac.uk/events/2006-04-10-12/|OSSW conference]] at Oxford) * Tackling this issue by considering all angles is a complex problem that would need external funding to sponsor a project team. We recommend approaching the [[http://www.jisc.ac.uk/index.cfm?name=about|JISC]] and OUCS would be happy to support this effort. * '''NEW''' With the current focus on climate change advise in this area is becoming more readily available. Here is a good source from Computing magasine: http://www.computing.co.uk/greencomputing * Tools are available to support PC users in lower energy use: * With XP/2003, a utility is present called powercfg. This allows a finer granuality of control of the power options, which include the option to throttle the processor when not in use. It would be interesting to attach a power meter to a PC and try out various options here for PC in 'standard' use. Change one option a day and see that happens. More info here: http://support.microsoft.com/kb/324347 * User help tool for PC Energy management: http://www.localcooling.com/ == References == 1. OUCE report: ''Computing the costs'' (not currently available via the web). 1. [[attachment:IT_Power_Saving.pdf]] 1. http://www.sun.com/emrkt/trycoolthreads/products.jsp 1. OUCE meeting actions: https://wiki.oucs.ox.ac.uk/oucs/fbtt06 1. PC energy usage calculator: http://www.energystar.gov/index.cfm?c=power_mgt.pr_pm_step1 1. Power usage per PC model: http://www.enviroquiet.com/ 1. Environmental champions at Oxford: http://www.admin.ox.ac.uk/estates/environment/dec.shtml 1. Recycling of computing equipment: http://www.wasteonline.org.uk/resources/InformationSheets/ComputerRecyclersRefurbishers.htm 1. Simulation of cooperative vs greedy resource consumption: http://ccl.northwestern.edu/netlogo/models/Cooperation (managing an 'economy' to encourage cooperation/ sustainable consumption is difficult!) 1. Computing magasine: http://www.computing.co.uk/greencomputing == Investigation notes == === Timing of software updates === * From PCPro (May 2005): {{attachment:Security.jpg}} * BL: Only if you're doing things to them overnight. * BL: Most likely the standard Windows updates are being used for patching. I don't think there is an option to force a shutdown within a particular time frame. I think it's either "force shutdown no matter what" or "give user option to delay". If overnight, the first can generally be used; if during the day then option 2. is the nicer way to go. * BL: Going the overnight route tends to ensure that the update gets applied more quickly as if you're being nice during the day, then even though the prompt message start to appear c. every 10-15 minutes, people will keep putting off the reboot, sometimes for more than one day (I know because I'm one of them!) * BL: In other words, yes there's a case for waking machines up overnight and patching then shutting down (possibly what we'd look to do in OUCS in the longer term). * BL: These can require a reboot, which people tend to find inconvenient. In reality it probably requires 5 minutes inconvenience, but... overnight is probably preferred for installation from this point of view. * BL: Generally it probably is more convenient/sensible to update security patches etc. overnight. However the happy medium is probably to have a system that wakes all the PCs up at some time in the night, patches (or allows patches to install), then shuts them all down again. * BL: Something like Altiris will allow scheduled wake-ups and shutdowns (although it's a bit of a sledgehammer to crack a nut if not using it for other things as well). There are probably free tools which when combined with some scripts would achieve much the same result. * IS: all HFS backups occur overnight. Desktops typically have a once a week backup slot (OUCE's is Tuesday evening 18:00 - 24:00 ). So, switching these desktops off would cause the automatic backup to be missed. The possible workarounds are: * Desktops store all useful data on a local server and this gets backed up nightly. * Leave those PCs on that are registered for HFS backup ( only 9 ) on the night in question only ( Tuesday night ). * Ignore the overnight backups and just run manual backups during the day. === Rebooting and hard disk damage === * From PCPro (May 2005): {{attachment:Security.jpg}} * BL: Not that we're aware of. The lecture rooms and Help Centre are probably a very good test case here. The machines get used every week day, get started up in the mornings and shut down every night for the majority of the year. They may also get shut down in between courses. We've never seen any problems that can be attributed to excessive starting up and shutting down. Technically it is possible that there is a difference and it is "better" to leave machines on all of the time (I don't actually know), but in practice I think that any difference is so slight that it isn't noticeable over the normal lifetime of a machine (so this is irrelevant). === Remote access === * BL: Because the machine has to be on to do this. And unlike patching, this is unpredicatable, and you cannot simply wake a machine for 2 hours in the middle of the night so that staff can remote desktop in! I don't know whether this is used in OUCE, but it would be a problem in OUCS for example. Not that there aren't ways around it, but they tend to require something being able to trigger something on an unscheduled basis - more difficult. * BL:Well, we've thrown this one around a bit (theoretically) as we know we would have this problem within OUCS and it would be good to have a solution here as well (not that we can necessarily do anything fast but we may be able to incorporate it into the design of the ECE). * BL: We think it should be possible with some appropriate scripting/programming, one or two (probably free) tools (or Altiris) and the odd web page to offer an interface whereby someone could wake their machine up when they needed it. We're also just starting to look at scripts to shut down machines automatically if they've been idle for a given length of time. There will be other things that other people are trying as well (there's been some recent related discussion on the US Windows higher-ed list). I know also that there are group policy related power settings that can be used, although they're not necessarily 100% reliable (yet). These will tend to require Active Directory, but OUCE are using AD already. * BL: Overall we don't have many tried and tested solutions, but we know of a lot of things which, if put together, should give reasonable solutions to most of these problems. * HN: Is remote access to a PC really required. Are users using their PC as a web server when they could be using a dedicated machine i.e. moving their files up to a web server.