Slashdot Mirror
Power Management and Networks?
ChamaraG asks: "Do you enable power management in your desktop PCs, and have you had any problems with networking after enabling power management (problems like losing open network connections, network using applications hanging after resuming from low power states, etc)? To clarify, by desktop PCs I mean PCs compliant with ACPI and Wake-On-LAN and capable of resuming from low power states in a few seconds, so that waking up time is not an issue. I am interested in the energy efficiency of networks and networked devices and I would like hear of problems that you might have had. Some applications I have tested will disable power management settings, presumably in order to maintain network connectivity. Surveys by the Lawrence Berkeley National Laboratory show that less than 5% of desktop PCs in offices are in low power states at night (36% - off, 60% - on). So, do you enable or disable power management in your PCs? If power management is disabled, what prompted you to do so and what would make you enable power management? What connectivity related problems did you encounter after enabling power management?"
Considering modern ATX power supplies common to most desktop PCs, wouldn't "off" be considered a "low power state"? (ATX power supplies use a small amount of current even when the computer is off, unless the power supply is plugged.)
In that case, the stats will be that roughly 40% of desktop PCs are in a low power state at night.
I never use power management, except for turning off my monitor after about an hour. There are too many applications that freak out during powersave, especially ones that try to access the hard disk for various reasons. Think about it: how often do programs you write have extra code in them for handling power-save modes? Probably not many. So when your application tries to get info to/from a powered off hard disk, swap partition, etc., it many times freaks out (especially 3D accelerated games in the middle of eating dinner). I used Windows for many years, and this means lockups and crashes. Eventually I stopped using power-save options altogether, in favor of stability. I've never tried any power-save options on Linux, though. I've just been conditioned not to use them I guess. But the most important thing is the monitor; no programs freak out when there is no monitor, and the monitor sucks up the most juice anyway. So poweroff monitor after an hour is all I ever do.
For XP Home edition on Dell computer: No deep hibernation allowed (computer won't wake up), no Ethernet card power down allowed (always loses the connection and has to be restarted). Everything else seems to do ok. Monitor going to sleep is good (saves screen, saves heat, better screensaver than some silly graphic acrobatics and no wasting cpu cycles). Hard drive spin down is ok as long as it doesn't happen too fast (at least one hour of inactivity). Of course, running SETI or some other background community software means your computer can't sleep.
FYI, the statistics in the report do not include home computers.
The NSA: The only part of the US government that actually listens.
If power management is disabled, what prompted you to do so?
Some of them are servers. The rest run Folding@Home.
and what would make you enable power management?
Being completely unable to afford not to. We've got quite a ways to go before energy becomes that expensive.
I hate idle computers, and by definition a computer in power-saving mode is idle.
Random and weird software I've written.
We schedule our WSUS (MS Software update services) to run at night so as not to bother the user. We also run quite a few machines with Deep Freeze, and they require a "Maintainence Mode" in order to unfreeze and apply updates.
My Staff are also becoming addicted to running Remote Desktop, so they won't let us control when the machines are turned off.
Most new staff ask me what they should do with their PC at night, and I always feel guilty when I tell them that it doesn't matter to me, and they might just leave it on. (Even if they turn it off, it auto-ons itself at 4:00 a.m.)
I'm always using my CPU, so I don't want it to go into low power mode. I support the Folding At Home and IBM's World Community Grid projects.
Even if I wasn't, I'm often still using my CPU for keeping Azereus running (Fedora distros).
I don't want it going offline. I want it doing my bidding full time.
Unitarian Church: Freethinkers Congregate!
I grew up with MS DOS 3.3 and progressed until Widoze 98 and then switched to GNU/Linux. I had a period whith 98 where I wanted to test power saving, but it always made the programs and network unstable or useless. So I don't use it... besides the few seconds that it takes a computer to "come back to life" and then to be rebooted is not worth my time when I need my computer... and with encrypted wireless pcmcia it is utterly useless.
:-)
But there are some good in the screen controll, but I need the screen to be on all the time else I can't see what happens or who writes on ICQ/MSN... just hearing a sound is annoying, if you know what I mean... need to know what is going on "out there"
Vegetarians eat Vegetables, Humanitarians frighten me...
At my job, we have all the servers & workstations configured to run 24/7, for several reasons.
On each computer, scheduled jobs (cronjobs on Linux & OSX, scheduled tasks on Windows) do things like backups, updates, antivirus & antispyware scans, etc. Making these things happen automatically and outside of business hours makes life much easier for the people that have to maintain all these systems (i.e. me). The main downside is that occasionally an operating system will force an unattended, unprompted reboot, causing the owner of that workstation to lose work. This is annoying, but we warn people in advance that it could happen, and they need to save their work before going home; the alternative -- manually finding out about & doing these updates machine by machine -- just isn't tenable.
The other benefit though is that we have SNMP monitoring of all hosts throughout the day. If a machine is having a hardware problem -- say, the printer is running out of toner, or someone's hard drive is filling up -- we get alert emails about it so that the problem can be fixed. This also allows us to know what machines are down at any given time, and have a reasonably small window indicating when it went down and when it came back up; this can help narrow down time frames for events like office power outages or, should it happen, the theft of a workstation.
I suppose we could get some of these benefis while also providing nightly downtime, but the benefits of having continual monitoring & maintenance are strong enough that there hasn't been a lot of call for it -- and when someone brings the idea up, the proposal usually gets shot down by the people managing the network.
If it were possible to bring this equipment into some kind of standby mode where it was still possible to do basic network tests (continually pingable, occasionally query for things like disc space, once-nightly wake up for updates & virus/spyware scans, etc) then maybe the idea would fly, but as things are now, there doesn't seem to be a good way to get these benefits other than by just leaving everything on all the time.
If nothing else, at least the monitors get turned off when they go idle...
DO NOT LEAVE IT IS NOT REAL
i could be wrong, but i'm pretty sure magic packets don't route: they broadcast. this'd reduce the usefulness considerably in a corporate environment...
Certainly there is too much software that's either not sleep-savvy or just sleep-broken. But that doesn't mean you can't do anything.
:)
First, set the displays to sleep. That's easy and saves plenty of energy, especially with CRT's..
Next, set the machines to sleep that can do it well. I set all the Mac workstations to sleep, they handle things well. Linux and Windows machines depend on the hardware and how well it sleeps.
Keep this in mind for your ongoing infrastructure plans. The more you centralize data on file servers or users on terminal servers the more you can sleep at night (no need for workstatioon backups, do a WoL for automated updates).
And request your vendors to support sleep. Apple needs a 'wake for cron jobs' checkbox in the Energy Saver control panel. I could sleep another dozen machines if they gave me that.
Then think of the amount of energy resources that could be saved if everybody did this. I must number in the billions of barrels of oil. Maybe Homeland Security should be in on it.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Do hard drives actually spin down under XP? (or NT or 2K or a Linux?) I've had that power saving feature set on my laptop and desktops for years. I don't think that I've noticed a hard drive get a chance to spin down in anything more recent than the late 90's and Windows 98. There are so many background tasks that seem to prevent that bit of power savings...
Jesus, completely forgot MS took "democracy" and other words out of the chinese internet, ala newspeak 1.0!
Sarcasm should not be allowed to be this realistic, please tell them all to cut it out.
The first rule of USENET is you do not talk about USENET.
I've been putting my desktop to S3 sleep every night for the last year or so. Overall, it works really well.
My machine is an Intel 845BEG desktop board running FreeBSD 5.4-STABLE. The ACPI support in FreeBSD is quite good. I bought the motherboard specifically to be able to suspend it. I reasoned that because Intel pushes ACPI, they would have a good implementation in their BIOS.
The one application I have problems with is kmail. After a resume, it will think it lost connectivity to the pop server and sometimes hang. I just kill and restart it. That's much faster than having to re-open all my various emacs buffers. And of course my ssh sessions are dropped. Overall, I'd say its just the same as suspending a laptop.
I have one add-in PCI card that doesn't understand sleep mode (a Comtrol Rocketport serial card). I have a script to unload the driver before I put the machine to sleep. The I re-load the driver when the machine wakes up.