Over 65% Power Reduction on my Ubuntu Server!

Wow!

I started using the Beta MicroMiser software this morning. I have installed it on the following interesting computers at my home. (They all run Ubuntu. The server is on 8.10 Server edition the others are on 9.04 desktop all 32bit):

  • Lobsang – Desktop PC Intel Core2 8400,
  • Angua – Laptop PC Intel Core2 Mobile T5500,
  • Vimes – Home Server VIA C7

I checked with MiserWare that they were OK for me to report some of the stats and they said it was OK. So, if you run the daemon on your machines, look in /var/log/syslog for lines like the following.

Lobsang reports about once per hour with:


May 13 17:09:32 Lobsang micromiser[7571]: Estimated energy saved since MicroMiser start: 0.100145kWh (20.70%)
May 13 18:09:32 Lobsang micromiser[7571]: Estimated energy saved since MicroMiser start: 0.104223kWh (19.07%)

which is not bad. It’s a fairly hungry processor so ~20% saving is certainly not to be sniffed at.

Angua reports :


May 13 19:05:51 Angua micromiser[2496]: Estimated energy saved since MicroMiser start: 0.000124kWh (33.9%)

Which is even better percentage-wise and means my battery should last a bit longer when I am out and about.

And finally Vimes, my server that is on 24/7 reports:


May 13 17:12:35 vimes micromiser[14718]: Estimated energy saved since MicroMiser start: 0.315464kWh (65.19%)
May 13 18:12:35 vimes micromiser[14718]: Estimated energy saved since MicroMiser start: 0.356811kWh (65.26%)

Which is pretty astonishing frankly. Now these figures are for just the processor but even so, that VIA C7 is described as drawing about 7W under normal load. If I have more than halved that over the lifetime of the device this will not be insignificant.

Really fascinating stuff.

If you want to try this at home on your computers, please leave a comment below and I will send you an invite. The trial runs until the end of June.

Tags: ,

43 Comments

  • hardran3 says:

    This looks like an amazing program! I wonder if it would work on my Acer Aspire One?

  • Alan Bell says:

    have you measured it independently? i.e. at the plug? I have a power meter and I will give it a go later.

    • Alan Lord says:

      No I haven’t.

      And just to be clear, this is just the processor we are talking about; not the whole computer that is being affected.

      @KaffeLars and hardran3,

      Done.

  • Chris Harland says:

    Looks like a sweet program I’d love to give it a try.

  • Chris Irwin says:

    How is the savings calculated? Based on your previous governor (ondemand, etc) or based on standard useage (performance)?

    • Alan Lord says:

      I don’t have any idea yet – only came across it this morning. I’m sure we’ll get more detail as the product develops. There is some IP in there with regard to the algoritms they are using, but the app itself runs in userland. You have to disable other frequency scaling apps before running this one.

      I’ve added you to their list too.

      • Chris Irwin says:

        Thanks for the invite, but I have to decline. While my primary machine is a laptop, I currently spend about 90-100% in C4 state (according to powertop) as I’m idle typically. I would imagine power under load would be identical, so I won’t discuss that.

        I’d like to see your power usage with and without this utility. Consider that I can save an additional 3-4W by simply turning my brightness down. To run a closed-source utility to save maybe 1W of power? I’m not sure I can do that. If they were writing a new scaling governor for the kernel or making their userspace algorithm public, I would gladly use it and send them any profile information they needed.

        • Alan Lord says:

          @Chris,

          It’s no problem. It’s your choice of course. I will collate some of these comments tomorrow and feed them back.

          I have enquired about the closed nature of the code, and most of the wrappers are BSD licensed in Python but the algorithms they use are closed.

  • joel says:

    i’d like to give it a try. thanks

  • yungchin says:

    Thanks again for sharing an invitation with me. In the end I couldn’t agree with the license terms I’m afraid… In fact, I think under the license terms for the beta you’re not even allowed to share the information you provide in this post! That was a bit too much for me…

    • Alan Lord says:

      That’s your choice of course. I thought about it for a while too. But I had a lot of faith/respect for the person who introduced it to me.

      As for the content of this post, that was fine. I checked with them first. They are more concerned about “unofficial” benchmarking apparently. As the product is still in Beta I can understand why I think.

      Cheers.

      • Chris Irwin says:

        Would they let you post information from a kill-a-watt or even powertop publicly? or just the numbers their software provided to tell you how great it is?

        • Alan Lord says:

          I can’t see how they could stop you from making independent reviews of various technologies. But IANAL πŸ™‚

          • yungchin says:

            Well, as they write in their License Agreement “any information learned or discovered by You about the performance of the Software in the course of use under this License constitutes proprietary trade secret information” – so getting the software means you’re bound by a NDA. I’m not a lawyer either, though.

          • Alan Lord says:

            @yungchin,

            I have received an email from Miserware regarding their license terms. They have updated them and made them a little less onerous:

            Several folks have asked about benchmarking against other power management software. With regard to benchmarking, we want to clarify the intent of the license. Our intent was not to preclude benchmarking altogether, but to ensure the measurement methodology is fair to all parties. More precisely, our lawyers told us to paste both permissions together:

            “You are hereby authorized to disclose information regarding the performance of the MicroMiser software, (i) provided that such information is provided to you in a MicroMiser software report, or (ii) provided that such information is obtained using techniques approved in writing by MiserWare.”

            The changes are retrospective to existing accounts as well as all new sign-ups.

  • Joakim Berglund says:

    I’d really like to give this a try too. My Always-On Home Server could use something like this… Thanks!

  • Tom says:

    Can you send me an invite?

  • Bambitroll says:

    I noticed that one has to remove powernowd which is already helping using less energy. So the real saving figures should be computed as the difference between the machine running standard Ubuntu with powernowd and Ubuntu without powernowd but with microniser.
    Don’t you think?

  • Bambitroll says:

    The “Estimated energy saved” figure is computed against a computer running no power saving software at all isn’t it?
    By default, Ubuntu runs powernowd which helps saving energy.
    So the real interesting comparison would be what benefits we get with micromiser compared to powernowd…
    Or am I just completely lost here? πŸ™‚

  • Kiran. says:

    Hello Alan,

    If you still have invites left please send me one. I run dual core machine (up 24/7) and two laptops at home for personal use.

  • Chris Maier says:

    Sounds really interesting, I’d like to give it a try too! Thanks πŸ™‚

  • iggykoopa says:

    you can also try an app I wrote for powermanagement here http://code.google.com/p/wattospm/ . It works pretty well so far and does a lot more than just your cpu.

  • Alan Lord says:

    So everyone who asked for an invite should have received them. If you haven’t please let me know and I’ll check.

    I have not been told to stop asking for new participants so I’m assuming they have more space. In fact, you can – once registered – start inviting your own colleagues too.

  • Alan Lord says:

    @Kiran, I just checked and it should have been sent. I can’t enter it on the system again.

    “This email address has already been invited to the MicroMiser beta. If you believe this is in error, please contact us.”

    Can you check any spam buckets?

  • Kiran. says:

    Alan,

    OOps received it, its in my Thunderbird’s Junk Email (dunno why). Sorry for any trouble and thanks!

  • kendon says:

    send me an invite please, laptop’s battery has become old… πŸ™

    have you actually measured the energy consumption of your devices? the output of the tool sounds promising, but a command line is patient, if you know what i mean…

  • Alan Lord says:

    Every one should have had their invitation.

    @Kendon, no I haven’t. I don’t have a power meter. It would be a good thing to do though but do recall that this software is tuning the processor only. So if you have “thirsty” hard disks, inefficient power supplies etc, then these are other areas that can also help to reduce consumption significantly.

    But, as the saying goes “Every Little Helps”…

  • Pana says:

    Interesting. I would like to try it if you have any invitations left. Thanks in advance.

  • Chris Phillips says:

    I’d love an invite if you have any left. Thanks!

  • Alan Lord says:

    Invites all sent out. Please shout if you want any more.

    I think Miserware are running the beta trial until at least the 30th June.

  • David Mohammed says:

    Alan – any updates on my request for a invite (from your “Power Saving for Linux” article)? TIA

    • Alan Lord says:

      Hi David, please check your spam filters. The invite was sent out yesterday to the email address you supplied.

      Thanks

      • David Mohammed says:

        Alan – I think that my email account sees you as ‘junk’ – not my fault honest! – the spam filter was set to autodelete. I don’t know your email to put it into the safe-senders. Any chance you can send to this revised gmail – alternatively can you resend to hotmail – I’ve tried to change it to not autodelete. TIA

        • Alan Lord says:

          Hi David, no problem. I’ve sent another invite to your gmail account. Hope that one arrives.

          • David Mohammed says:

            Alan – it has many thanks. It claims its saving a steady 33%. I’m intrigued. I’ll leave my laptop on to see if I get 4 hrs instead of the usual 3. Many thanks for your help.

  • UxBoD says:

    Hi,

    I certainly would like to be added to the trial. I have three servers in my home office ranging from a Asterisk Mini-ITX server (low wattage anyway) up to a Dual CPU Opteron server. All servers are on 24×7 so would love to reduce my power consumption a bit πŸ™‚

    Regards,

  • […] interesting. Even better, Alan wrote a follow-up post at The Open Sourcerer which shows that on his Ubuntu Server system the software estimates its own […]

  • […] well over a week now since I started using the Miserware MicroMiser software. I have it installed on all the Ubuntu PCs we have at home and on […]

  • Ushao says:

    I’m curious about this on my own server. It’s on some older hardware so it’s not as efficient as the newer stuff. I’d like an invite if possible.

  • Weazle says:

    I also would like an invitation to try the MiserWare software, thanks in advance!

  • x-quadraht says:

    Well, as far as I can see it’s not doing anything other than switching the CPU to lower performance states with less power consumption if there is less system load. That is something the linux kernel already does on its own if you use the “ondemand” cpu frequency governor which is activated per default on many distributions.
    I expect that the “power saved since micromiser start” is nearly equal to what the linux kernel saves if left to its own. I would be very interested in a benchmark which compares a system fine-tuned with powertop and a micromiser system (also fine-tuned with powertop before to make equal starting conditions)…

  • […] by some of Alan’s posts at the Open Sourcerer, I’ve been conducting an experiment in recent weeks with the […]

Leave a Reply

XHTML: You can use these tags:
<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>