Slashdot Mirror

← Back to Stories (view on slashdot.org)

How Facebook Is Saving Power By 10-15% Through Better Load Balancing

Posted by timothy on from the 10-percent-at-web-scale dept.

An anonymous reader writes Facebook today revealed details about Autoscale, a system for power-efficient load balancing that has been rolled out to production clusters in its data centers. The company says it has "demonstrated significant energy savings." For those who don't know, load balancing refers to distributing workloads across multiple computing resources, in this case servers. The goal is to optimize resource use, which can mean different things depending on the task at hand.

54 comments

  1. 10%-20%? by Anonymous Coward · · Score: 0

    Somebody's lying.

  2. They could save 100% by Anonymous Coward · · Score: 5, Funny

    Just turn it off.

    1. Re:They could save 100% by Jane+Q.+Public · · Score: 1

      I agree. Facebook should have no power at all.

  3. to sum it up by roman_mir · · Score: 3, Informative

    to sum it up, if a FB server is idle it consumes 60 watts, if CPU is minimally utilised it consumes 130 watts and if it's utilised more it consumes 150 watts.

    Instead of round robin use an algorithm that pushes requests to the servers that are already processing other requests, thus allowing many CPUs to remain at 60 watts, while some CPUs to hit 150 watts of power consumption and so instead of doubling or almost trippling power consumption of all servers due to round robin distribution of requests, tripple power consumption of fewer CPUs and let many CPUs to stay at 60 watts.

    Sure, it's an interesting thing to optimise, but unless you are running dozens or maybe hundreds and even thousands of servers in a data centre you won't care about this much at all.

    --
    You can't handle the truth.
    1. Re:to sum it up by Anonymous Coward · · Score: 0

      Or if you're Intel or AMD making millions of CPUs, you think about how to do your systems right.

      Or even Comcast or Verizon. Oh wait, they don't have to pay for the power their boxes use.

    2. Re:to sum it up by roman_mir · · Score: 1

      Or if you're Intel or AMD making millions of CPUs, you think about how to do your systems right.

      - what does that have anything to do with the question at hand? Nothing at all, CPUs will use more power when they actually have to process something and less when they idle, what are you talking about? The only interesting question here may be that CPUs need to know about each other and distribute the load in a way that would reduce power consumption while still providing the necessary processing, but this algorithm does that exactly and it's probably simpler than setting up CPUs in such a way that they would do that work on their own without knowledge of the context of the requests that are being executed!

      --
      You can't handle the truth.
    3. Re:to sum it up by nabsltd · · Score: 1

      Or if you're Intel or AMD making millions of CPUs, you think about how to do your systems right.

      - what does that have anything to do with the question at hand?

      Well, perhaps somebody like Facebook who buys at least thousands (if not hundreds of thousands) of CPUs might go to Intel or AMD and have them custom-build (or at least custom tune) a CPU for their use case.

      Also, looking hard at the actual usage might help...lower clock speed CPUs with more cores and lower TDP vs. faster CPUs with fewer cores and higher TDP doesn't have a set answer. Perhaps the latter are better because they finish the job quicker and get back to idle, but maybe the former works better because it can handle more user requests per watt when busy. I have no idea, but answering these kinds of questions might save more power than just making sure idle CPUs stay idle for as long as possible.

    4. Re: to sum it up by Anonymous Coward · · Score: 0

      The CPUs are one of the easy to recognize costs to a system, though I could also concur with concerns about data access and even cooling, if we wante. to digress into that, however the point being made was that the makers of chips like Intel and AMD have their own role to play in this. Why even the ability to prioritize some systems could be said to be in their hands. After all, they are at the point where they can make the power management signals work most easily.

      Certainly a lot easier than us at that bottom end of the product chain. Not like I can adjust the mains power like I could with an incandescent bulb .

    5. Re:to sum it up by Anonymous Coward · · Score: 0

      So another way to look at it is that facebook has too many computers if they have CPUs actually idle....

      Or they were wasting 10-15% power....

    6. Re:to sum it up by sociocapitalist · · Score: 2

      to sum it up, if a FB server is idle it consumes 60 watts, if CPU is minimally utilised it consumes 130 watts and if it's utilised more it consumes 150 watts.

      Instead of round robin use an algorithm that pushes requests to the servers that are already processing other requests, thus allowing many CPUs to remain at 60 watts, while some CPUs to hit 150 watts of power consumption and so instead of doubling or almost trippling power consumption of all servers due to round robin distribution of requests, tripple power consumption of fewer CPUs and let many CPUs to stay at 60 watts.

      Sure, it's an interesting thing to optimise, but unless you are running dozens or maybe hundreds and even thousands of servers in a data centre you won't care about this much at all.

      Some of us do actually run hundreds or thousands of web servers so it is actually interesting to us.

      Also, I think the idea is not only applicable to web servers. I'm not an expert in this field but I would think the power consumption difference is due to dynamic frequency scaling both by direct consumption and by subsequent heat generation.

      --
      blindly antisocialist = antisocial
  4. Switch off servers? by manu0601 · · Score: 2

    TFA is vague on that point: do they switch off some server during idle hours?

    Such a practice seems good for power consumption, but we have to account the fact that switching on and off shortens hardware lifetime: it creates temperature stress, and we all know that electronics most often die at power on time. Hence what looks like a power saving may hide bigger costs (either financial or environmental) for hardware replacement.

    1. Re:Switch off servers? by Johnny+Mnemonic · · Score: 1

      No, they are simply letting the CPU util go to 0% (+ whatever necessary for OS etc). But the hosts are still awake and available. Another advantage is that the load can be instantly added back, whereas if they actually turned the machines off they'd have to wait for boot time, so the reaction to capacity shifts wouldn't be as fast.

      --

      --
      $tar -xvf .sig.tar
    2. Re:Switch off servers? by manu0601 · · Score: 2

      I see the value of having a pool of idle servers ready for request peaks. But the graphs in TFA shows they have huge daily variations, hence it could make sense to switch off a fraction of idle servers.

    3. Re:Switch off servers? by Lennie · · Score: 1

      I wouldn't be surprised if you turn off the machines there is also a larger chance of failure.

      So when you try to turn it on, it wouldn't turn on or some disk would have not spun up.

      But I've never done the numbers if this is actually true.

      --
      New things are always on the horizon
    4. Re:Switch off servers? by Dogers · · Score: 1

      I quite like this artificial demo of VMwares DPM functionality:
      http://www.youtube.com/watch?v...

      --
      I am a viral sig. Please copy me and help me spread. Thank you.
    5. Re:Switch off servers? by profplump · · Score: 1

      Or it could be that the power-cycle stresses actually aren't a big factor, or aren't a big factor given the expected lifecycle of the device, or are a big factor but not big enough to offset the savings, and so it would make perfect sense to turn them off any time you're fairly sure it's safe to add the delay of a boot cycle. It's also possible that reducing power usage might be worth more than the pure cost of power, as it might reduce say, expected future power costs or installation costs or any of 100 other things.

      We get it, there are lots of potentially complicating issues. But it's silly to talk about how factor B might overwhelm factor A when you don't have numbers for factor B.

    6. Re:Switch off servers? by manu0601 · · Score: 1

      it's silly to talk about how factor B might overwhelm factor A when you don't have numbers for factor B.

      Nor for factor A!

  5. Should we care? by penguinoid · · Score: 3, Insightful

    Is this a case of "Facebook was being obliviously wasteful" or a case of "Facebook discovers way to increase efficiency"? I'm guessing it's the former.

    --
    Don't waste your vote! Vote for whoever you want, unless you live in a swing state it won't matter anyways
    1. Re:Should we care? by wonkey_monkey · · Score: 1

      Don't they both amount to the same thing?

      --
      systemd is Roko's Basilisk.
    2. Re:Should we care? by penguinoid · · Score: 3, Insightful

      The first is like increasing fuel economy by designing a better engine, the second is like increasing your own fuel economy by patching that leaky fuel tank. Even I know how to turn off or put my computer into sleep mode to save energy -- will anyone but Facebook gain anything from this?

      --
      Don't waste your vote! Vote for whoever you want, unless you live in a swing state it won't matter anyways
    3. Re:Should we care? by Anonymous Coward · · Score: 0

      Facebook my have physical servers they can't easily turn off completely, but think of companies that use VMs in a cloud provider. They could decrease/increase number of VMs. That may be substantial savings.

    4. Re:Should we care? by Anonymous Coward · · Score: 0

      By definition Facebook is wasteful.

  6. We all could save energy... by MS · · Score: 2, Funny

    Not only Facebook, but the end-users also could save a lot of electricity by not using Facebook at all. People should get out and have a real social life.

    1. Re:We all could save energy... by Anonymous Coward · · Score: 0

      Not only Facebook, but the end-users also could save a lot of electricity by not using Facebook at all. People should get out and have a real social life.

      There is research on this (was even a story on Slashdot a while back) that actually prove the opposite correlation of what you are implying. Active Facebook users have a more active "real" social life than non-Facebook users. It is also interesting that you most often see this argument on sites like Slashdot with a demographic not known as the most outgoing social ones.

    2. Re:We all could save energy... by umghhh · · Score: 1
      Indeed but I still save power but not going out and not using FB.

      Besides I tried the world outside and after all the years I realized that it was not worth it.

    3. Re:We all could save energy... by Anonymous Coward · · Score: 0

      There is research on this (was even a story on Slashdot a while back) that actually prove the opposite correlation of what you are implying.

      Most such research is biased, shortsighted, or takes subjective criteria into account and reaches arbitrary conclusions based on that. But I'll assume it wasn't.

      Active Facebook users have a more active "real" social life than non-Facebook users.

      It's unlikely that that's because of Facebook; it would be true whether or not Facebook existed.

  7. I'm sure by scorp1us · · Score: 1

    There's some savings to be had by, if you have a geographically distributed system across time zones, moving loads to lower commercial rates based on time zone.

    For those that don't know, commercial rates vary, and spike at peak demand time (~14:00) Moving peak load by forward or back 2 time zones would move you out of peak rates.

    --
    Slashdot's rate-of-post filter: Preventing you from posting too many great ideas at once.
    1. Re:I'm sure by Lennie · · Score: 1

      You can't do this for the webservers, because you build these datacenters to be closer to the users. To improve latency.

      So you are not going to increase latency to save power.

      I believe I heared someone from Google mention they do this for certain batch jobs, but I could be mistaken.

      The problem might be: you can only move such workloads if you have the data in the other datacenter too and the data is up to date enough.

      --
      New things are always on the horizon
  8. Re:irst Post ! by Jane+Q.+Public · · Score: 1

    "It's not _our tube, it's _is tube!" -- The Giant Rat of Sumatra

  9. It does not matter by Anonymous Coward · · Score: 0

    The more you save, the more you think you can consume. In the end, it's the same.

  10. this is slashdot 2014 by Osgeld · · Score: 1

    "those who don't know, load balancing refers to..."

    no shit sherlock

  11. Only two... by kefalonia · · Score: 1

    ... comments threads after 24 hours, on slashdot? Now we know which keyword turns off the slashdot crowd! :-P