Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
×
Cloud Google Virtualization

Google Offers Cheap Cloud Computing For Low-Priority Tasks 59

Posted by Soulskill from the guaranteed-7-7s-of-uptime dept.
jfruh writes: Much of the history of computing products and services involves getting people desperate for better performance and faster results to pay a premium to get what they want. But Google has a new beta service that's going in the other direction — offering cheap cloud computing services for customers who don't mind waiting. Jobs like data analytics, genomics, and simulation and modeling can require lots of computational power, but they can run periodically, can be interrupted, and can even keep going if one or more nodes they're using goes offline.
This discussion has been archived. No new comments can be posted.

Google Offers Cheap Cloud Computing For Low-Priority Tasks More

Google Offers Cheap Cloud Computing For Low-Priority Tasks

Comments Filter:

  • CCC4C (Score:4)

    by turkeydance ( 1266624 ) on Tuesday May 19, 2015 @06:50PM (#49730499)
    cheap cloud computing for customers. called it!

    • Re: (Score:3, Funny)

      by Anonymous Coward

      I want to found the Center for Cheap Collated Commodity-Clustered Civilian Cloud Computing for Cyclic Customer Code Collocation

    • Re: (Score:1)

      by Anonymous Coward

      This could be an inexpensive way to host the geocities mirror I made b4 they took it offline.

    • C4 .... oh wait, that one is taken.

    • the final 'c' is for cheapskate.

  • Spot Instances? (Score:1)

    by Anonymous Coward

    Preemptible VMs are the same as regular instances except for one key difference - they may be shut down at any time.

    It sounds like AWS's Spot Instances? Except for the fixed pricing.

    • Re:Spot Instances? (Score:5, Informative)

      by lgw ( 121541 ) on Tuesday May 19, 2015 @07:19PM (#49730691) Journal

      It sounds like AWS's Spot Instances? Except for the fixed pricing.

      Yup, it's their version. Forbes compares them. [forbes.com] The fixed price is nice on the Google side, but there's no 2-minute warning before termination on Google like you get on AWS, and AWS launched a new Spot Fleet [amazon.com] product the same day Google announced.

      Either way, you need to be doing the kind of work where you can lose VMs on short notice and keep going, but it's a very nice discount if you can.

      • Re:Spot Instances? (Score:5, Informative)

        by Fwipp ( 1473271 ) on Tuesday May 19, 2015 @08:01PM (#49730949)

        They do give you 30 seconds of warning: https://cloud.google.com/compu... [google.com]

        Compute Engine performs the following steps to preempt an instance:

        Compute Engine sends a preemption notice to the instance in the form of an ACPI G2 Soft Off signal.
        If the instance does not stop after 30 seconds, Compute Engine sends an ACPI G3 Mechanical Off signal to the operating system.
        Compute Engine transitions the instance to a TERMINATED state.

        So if you're able to persist your state in less than 30 seconds, just watch for SIGTERM and you should be golden. Otherwise, checkpoint frequently.

        • It goes without saying (although I am reminding people) that frequent checkpointing should be done as standard operating procedure.

      • Re: (Score:3)

        by jopsen ( 885607 )

        Either way, you need to be doing the kind of work where you can lose VMs on short notice and keep going, but it's a very nice discount if you can.

        The only problem is availability... Short of maybe database and legacy software... You shouldn't be writing distributed system that can't handle individual node failure..
        So the only thing holding this back is the fact that they don't promise availability and that they can take down all your nodes at once.

        I would argument one ought to run a percentage of ones servers as spot nodes... or preeamable VMs.

  • Going back to dumb terminals? (Score:2, Interesting)

    by Anonymous Coward

    In a way the cloud was big way back when. I remember in the late 70's my high school had a matrix dumb terminal tied to a couple college servers. It seems we are going back to that way of thinking. Use minimal hardware and let the power of the processing be handled elsewhere.

    • In a way the cloud was big way back when. I remember in the late 70's my high school had a matrix dumb terminal tied to a couple college servers. It seems we are going back to that way of thinking. Use minimal hardware and let the power of the processing be handled elsewhere.

      It is not about the processing power, it is about the control. However controls the data owns the data, regardless of law.

    • It's more like using maximum desktop / laptop / handheld hardware and then let even more computations happen on much more massive systems.

    • Re: (Score:2)

      by dave420 ( 699308 )
      Cloud computing is far more than just that. What happens when the school wanted more pupils than they had computer power for? Could they magically spin up a new computer out of thin air for the new users to use? Cloud computing is the combination of having your own mainframes/servers, and being able to add to their number/reduce them as close to instantly as physically possible. This lets small players have the same computing power at their fingertips as large companies, which is not the case with the s

  • FIRST POST! (Score:3, Funny)

    by Anonymous Coward on Tuesday May 19, 2015 @07:14PM (#49730651)

    If only I was not using the low-priority cloud compting..

  • Google has decided to triple the cost of your off site services. Please send a cheque for $10000 today. We'll keep your data nicely safe and locked away where you can get it as soon as you pay your new ransom I mean SAAS fee.
  • I have been programming since vi was considered advanced. I have to admit that features such as builtin debugging and autocompletion are very useful. I program in multiple languages and an IDE which understands the language I am using helps me greatly, especially when it knows about the library. I am more productive because of these tools.

    • "I have been programming since vi was considered advanced."

      We'll need more information to determine how long you have been programming, since you could have started moments before your post based on that criteria. VIM is still one of the best IDEs available (people mistakenly refer to it as an editor; that's like calling a car a windsheild; yes VIM has an editor, but it is much more than that.) Sure, you need to set it up with the right tools (e.g. tag files with cscope or ctags) but that is true of any m

    • You've posted to the wrong article.
      You should be using the Slashdot IDE to cut down on these errors and be more productive

  • Small comment on that last point:

    Any large-scale computing implementation must be able to deal with nodes going offline. For example, if typically you have one hour of downtime per machine per year, and are running a process on 10,000 nodes, then at any given time you have about a 64% chance of having at least one node down. If you have to restart the process on all 10,000 nodes, that's going to be a huge amount of wasted computing power.

  • State of the art from 1985? (Score:3)

    by awilden ( 110846 ) on Wednesday May 20, 2015 @12:12AM (#49732897)
    Back in the day, you could set your priority on a Cray XMP to 0.01 so that you got 100x the minutes you were allocated but that your code only ran when the machine was otherwise idle.

    • Indeed, load balancing and shifting is not new. It is, however, important and Google needs it to get maximum profitability out of their servers. To do so people need to know that it exists.

      Google did not claim there had not been other systems with similar load balancing. They simply explained what it was and how to use it. You know, the data you need to know if it is applicable to you.

    • Re: (Score:2)

      by dave420 ( 699308 )
      Could you get access to a second Cray for pennies? And a third? And a fourth? How much did the first Cray cost? That's the difference.
  • Or SETI@home. Now we can find out how much it costs to find each Little Green Man.
  • Looks like the perfect match to run systemd.

  • It sounds like this is modelled after the SETI@home project (which I think evolved into BOINC?), though maybe in reverse

    Berkeley: Hello, world, you're wasting CPU cycles, do you mind if we use them?

    Google: We've got a bunch of underutilized CPUs. Hello, World, how do you want to use them?

    I suspect the Google model would work better for data that can't be broken into parallel tasks as easily, but overall it sounds like both approaches are designed for similar things.

  • I like it (Score:4, Insightful)

    by tsotha ( 720379 ) on Wednesday May 20, 2015 @02:37AM (#49733453)
    This is a really good idea from a resource utilization standpoint. Reminds me a lot of the power grid - you have to have enough hardware to handle the peak load, but idle capacity just costs you money.

  • Amazon's spot instances are better in every way. Not only are they usually cheaper than Google's fixed prices, but you can run them for way more than 24 hours. I have a Spot Instance running for 4 months at about 1/25th the price of the on-demand instance, and way cheaper than Google's preemptible instance too.

    The limit of 24 hours seems to be designed to prevent people who want to run long-running tasks from using up the spare compute power on something like a VPN. That's fine with me; Amazon can have my b

Slashdot Top Deals

Stellar rays prove fibbing never pays. Embezzlement is another matter.

Close