Slashdot Mirror

← Back to Stories (view on slashdot.org)

Review: Google Compute Engine

Posted by samzenpus on from the silver-lining dept.

snydeq writes "InfoWorld's Peter Wayner takes an in-depth look at Google Compute Engine, the search giant's response to Amazon Web Services and Rackspace. 'If you want to build your own collection of Linux boxes, Google Compute Engine offers a nice, generic way to buy servers at what — depending on the size of compute instance you need — can be a great price. The most attractive feature will probably be the proximity to the other parts of the Google infrastructure,' Wayner writes, adding that Google Compute Engine is just one part of the Google APIs portal, a grand collection of 46 services. 'I suspect many developers will be most interested in using Google Compute Engine when they want to poll these Google databases fairly often. While I don't think you're guaranteed to be in the same zone as the service you want, you're still closer than when traveling across the generic Web.'"

13 of 60 comments (clear)

  1. How easy is it to leave? by Anonymous Coward · · Score: 5, Insightful

    You know, for when they shut it down?

  2. TFA doesn't give comparison by Taco+Cowboy · · Score: 4, Insightful

    What we truly need is a comparison chart of some kind, to show us what Google's offering is different from the others

    --
    Muchas Gracias, Señor Edward Snowden !
  3. I dunno... by Tastecicles · · Score: 2

    ...speaking for myself, and I don't know how many others reading might feel the same way, but $175 per machine per month (or however it's calculated, I only skimmed TFA) seems a little steep.

    Right this very minute I can fire up eight cores with combined RAM of something like 14.5GB - notwithstanding the fact that we're talking six machines (four laptops (two dual core - one Atom, the other AMD E350, and two P4/2.0 and 1.6GHz), two desktops (one P4/2.66, the other an AMD Athlon64/2.4GHz), it's very scalable depending on what I'm doing and how fast I want it.

    OK the Pentiums are *old* but they're still functional in a practical sense. I have older machines tasked for different things (a Thinkpad 760C running as a print server, for example) but I tend not to count or include them in my total computing power - they would suck up more in overhead than they would produce. Long gone are the days when a Pentium 120 made any sort of significant show in one of my clusters!

    About my only limitation is the number of available power sockets. Cue the arguments for cloud utility with the cost of domestic electricity supply. The key word here is domestic - I have complete control on how much power is used. My AMD laptop at full draw drinks 3.5kWh per week if I leave it on crunching 24/7. That's equivalent to £1.78 per month - or around 1/100 the cost of a single core on a cloud node.

    --
    Operation Guillotine is in effect.
    1. Re:I dunno... by viperidaenz · · Score: 4, Insightful

      But if you had nothing sitting around gathering dust, how much would it cost you to buy an equivalent amount of hardware? Keeping in mind depreciation, power , network access and infrastructure, and physical space. You also pay per hour, so if you don't need it one week, shut it all down and pay nothing. If you need something done very quick, pay for 100 cores for 1 hour instead of 1 core for 100 hours.

      It's not comparable to old laptops sitting in your attic.

    2. Re:I dunno... by RicktheBrick · · Score: 2

      IBM's Sequoia supercomputer has recently been declared the world's fastest supercomputer. It also seems to be energy efficient. The article stated it does 2 giga flops per watt. I would think that is better that any home or business computer. IBM is the main backer of World Community Grid and its computers are generating over 50,000 results a day. I would think that the cost of electricity for those results would be much greater than the cost if the results were generated on the Sequoia supercomputer. This is especially true since the results must be verified which means they are compared to another computer's results. This means each result is usually done by at least two computers doubling the cost for each result. Yet there are three members of wcg that produce over 20,000 results each day. If cloud computing is so great, I would think that one of the first things to end would be distributed computing because that seem to me to be the exact opposite of cloud computing.

    3. Re:I dunno... by jsdcnet · · Score: 4, Insightful

      But if you had nothing sitting around gathering dust, how much would it cost you to buy an equivalent amount of hardware? Keeping in mind depreciation, power , network access and infrastructure, and physical space. You also pay per hour, so if you don't need it one week, shut it all down and pay nothing. If you need something done very quick, pay for 100 cores for 1 hour instead of 1 core for 100 hours.

      At the personal level (not speaking of businesses here): AMD x8 FX-8150 3.6/4.2 GHz, 32GB RAM, 1TB HDD - DYI from all-new parts, no monitor - approx $750. Let's make it a full $1750 to allow for power delivered by "gilded electrons", "diamond optical fiber" supported internet access and a bouquet of flowers once a month for the "better half" to make it for the physical storage space. $1750 vs $175/month...10 month worth of VPS in Google's "compute" cloud (with 1 core with 4GB RAM).

      What about bandwidth? What about support? What about being able to be up and running on a new box in a few minutes when the old one takes a dump? You are totally missing the point of cloud services like this. It's not supposed to compete with your desktop machine.

      --
      no longer working for cnet
    4. Re:I dunno... by Anonymous Coward · · Score: 3, Insightful

      If your business needs move slow enough that pulling a physical box out of a cabinet somewhere, hooking it up, including peripherals, because you know you're going to need to at least configure it by hand far enough to get at least an automated installer kicked off, setting up remote access, is worth your time, six times, whenever you want to rebuild your utterly non-uniform cluster, then these services are not for you.

      For my time, I'd rather pay the TEN CENTS it costs to have Amazon do it for me.

    5. Re:I dunno... by DragonWriter · · Score: 2

      At the personal level (not speaking of businesses here): AMD x8 FX-8150 3.6/4.2 GHz, 32GB RAM, 1TB HDD - DYI from all-new parts, no monitor - approx $750. Let's make it a full $1750 to allow for power delivered by "gilded electrons", "diamond optical fiber" supported internet access and a bouquet of flowers once a month for the "better half" to make it for the physical storage space.
      $1750 vs $175/month...10 month worth of VPS in Google's "compute" cloud (with 1 core with 4GB RAM).

      If you are paying the premium that comes with a service that supports dynamic server provisioning and using it like you were just buying a VPS, you are doing it wrong.

      You should be paying for a VPS, which is different. You'll note that VPS's tend to quote there server prices per month. Compute Engine's are charged per hour. If you are using them flat across the month without variation, you aren't really understanding what they are for.

  4. Azure? by elabs · · Score: 2

    I can't belive Azure wasn't even mentioned. By some estimates it now holds more objects than Amazon (since iCloud and iTunes are hosted there).

  5. May work? by tanveer1979 · · Score: 4, Insightful

    This one may work.
    I was checking out the storage pricing, for example. Its 0.12/GB instead of 0.125 of amazon. If google does not do the "API fiasco", it will work. But as we all know, google and API is like chalk and cheese. Amazon is the king when it comes to accessibility.
    But I am glad, because this means more competition. So finally I can start using cloud storage, in conjunction with cloud server for hosting my blog etc., Heck, I can set up a machine in US, and get US specific content using VPN. No more looking for ssh services

    --
    My Aurora : http://www.youtube.com/watch?v=o91ZsGwJYyg
    FB : https://www.facebook.com/TanveersPhotography
  6. no way by Blymie · · Score: 3, Insightful

    Google will fail utterly and completely at this.

    Why?

    Customer service. They have a horrible, HORRIBLE customer service record. They just simply are unable to do customer service.... and this product needs it.

    1. Re:no way by Kotoku · · Score: 4, Informative

      You may have a point. Amazon is hard to top on that front. I had a billing question on EC2 at 2 a.m. and got an immediate response from a Seattle based employee and a service credit while I was doing adjustments to get my instances right. If their load balancers had a few more features I'd call it perfect.

      --
      AnimePapers.org: Anime Wallpapers Handled With Care
  7. Re:Azure... by CadentOrange · · Score: 2

    Because nobody gives two cents about Azure. Amazon dominate the cloud market and hence any new service will be compared against it.