Forgot your password?
typodupeerror
Google Cloud Technology

By the Numbers: How Google Compute Engine Stacks Up To Amazon EC2 76

Posted by samzenpus
from the head-to-head dept.
vu1986 writes "Google launched its EC2 rival, Google Compute Engine, last June, it set some high expectations. Sebastian Standil's team at Scalr put the cloud infrastructure service through its paces — and were pleasantly surprised at what they found. A note about our data: The benchmarks run to collect the data presented here were taken twice a day, over four days, then averaged. When a high variance was observed, we took note of it and present it here as intervals for which 80 percent of observed data points fall into."
This discussion has been archived. No new comments can be posted.

By the Numbers: How Google Compute Engine Stacks Up To Amazon EC2

Comments Filter:
  • by ethicalcannibal (1632871) on Sunday March 17, 2013 @11:49AM (#43197089)
    Just when you start relying on it, Amazon won't shut it down.
    • by mybecq (131456) on Sunday March 17, 2013 @11:53AM (#43197121)

      Just when you start relying on it, Amazon won't shut it down.

      Not intentionally, anyway.

      • Re: (Score:2, Interesting)

        by udachny (2454394)

        The question to Google is: what is the long term perspective here? Will this shut down if it doesn't generate some level of revenue/profit? What is that level?

        Is it possible to have a dynamically generated graph somewhere on Google that would show how far away is the break even point for this service, when will it become profitable (and this graph should be updated once in a while, every week or month to get a feel as to its long term prospect)? If it's a profit center on its own, at least there is a good c

    • Re: (Score:3, Interesting)

      by Necroman (61604)

      Sure it's fun to knock Google for shutting down services, but I believe most (if not all) of their shutdowns have always been free services they provide to consumers. I'm not aware of any paid Google service that has been shutdown. Though, Google has been known to drastically increase the cost of their services where it drives people away ( mapping [cnet.com] and AppEngine are 2 more recent examples, though they lowered the price of maps after a lot of people left).

      Google is trying to find services to hook people wi

      • by poetmatt (793785) on Sunday March 17, 2013 @01:22PM (#43197617) Journal

        paid service shutdown?

        how about picasa.

        google is not trying to hook people, but I still fail to understand why they shut down reader. It added a lot of value.

        • by Junta (36770) on Sunday March 17, 2013 @01:34PM (#43197675)

          google is a company. Companies don't really intrinsicaly about value provided to users as a rule. They care about the revenue they can get from their user activity. Reader porvides value, but Google seemingly doesn't see it as a revenue stream.

          Google isn't doing things out of the goodness of their hearts. A lot of companies give that impression as they ramp up, but inevitably a company will show it's capitalist nature, fail as a business, or both.

          • by poetmatt (793785)

            google has certainly started to slip, but they were one of the few companies that at least at first, did do things to benefit the users and not to benefit themselves. It was mutual. I don't think that's about some kind of capitalist nature/failing as a business. Those are not alternatives for any company, nor are they automatically how things occur.

      • Re: (Score:2, Informative)

        by Jason Earl (1894)

        Clearly you are not a Postini customer.

      • by Anonymous Coward

        The cost of Google maps drove us away. $10000/year minimum for doing some tests (a few prototypes) was just too much.

        But amazon isn't much better. They used to have an API you could use to search for books by ISBN numbers and similar. Not only there were several changes (like the complex authentication they added at some point), but now it's been retired, and you do have to subscribe to their new API which requires a lot of information and a credit card on file. Meanwhile, I can use a similar service by Goo

  • by Qwavel (733416) on Sunday March 17, 2013 @11:56AM (#43197143)

    I look forward to seeing Amazon and Google battle each other in providing Linux infrastructure. I know there are many excellent small providers, but no one has really come close to Amazon so far.

    • by houghi (78078) on Sunday March 17, 2013 @12:24PM (#43197295)

      So having Amazon and Google is good for competition, because they drown out the excellent small providers and thus having LESS options?

      Look at politics. Having only 2 things to select from is not competition. It is an illusion of competition.

      Yes, 2 is better then 1. But 3 is better then 2 and 20 is better then 3.

      • Re: (Score:1, Interesting)

        by udachny (2454394)

        Yes, 2 is better then 1. But 3 is better then 2 and 20 is better then 3.

        - only if they exist in the market that is free of regulations, that help one business to get a leg up against another. If it's the free market that provides competition, then yes. If it's done the way it was done in (for example) case of Standard Oil, then no. Standard Oil had plenty of competition and it was pushing prices down for 40 years before it was broken up. In these 40 years the company made very good profits, helping to make Rockefeller one of the richest people in the entire history of the wo

      • by Junta (36770) on Sunday March 17, 2013 @01:39PM (#43197695)

        Currently, EC2 pretty well dominates and stifle the small providers without any help at all.

        In fact, Google, Azure, VMware, IBM, HP and any other large providers coming into prominence may help the small providers. Currently, a lot of people beileve that hosting==EC2. If large competitors change the mindset to have customers realize there is a choice, that realization may have benefits. E.g. if a CIO directs a team to 'take everything to EC2', that's pretty much a guaranteed loss for the small provider. If CIO directs instead 'take everything to a hosting provider', that team then is empowered to allow more providers to compete for the business, even if the CIO mindset was changed only because of the big players.

      • by Teckla (630646)

        Yes, 2 is better then 1. But 3 is better then 2 and 20 is better then 3.

        At a glance, that sounds insightful, but . . . what about the paradox of choice [wikipedia.org]?

        For example, a lot of people think that having too many options is one of the reasons Linux on the desktop has not been much of a success.

        • Lots of people feel having a large choice of car manufacturers is the reason why motor cars have been a failure. Fortunately they are no longer in power in the USSR.
      • So having Amazon and Google is good for competition, because they drown out the excellent small providers and thus having LESS options?

        Look at politics. Having only 2 things to select from is not competition. It is an illusion of competition.

        Yes, 2 is better then 1. But 3 is better then 2 and 20 is better then 3.

        This remark is so very true. I'm so tired of having to choose political candidates based on one of two bases, welfare state or Christian Taliban. I wish I had a real choice that didn't include either.

        • I'm so tired of having to choose political candidates based on one of two bases, welfare state or Christian Taliban. I wish I had a real choice that didn't include either.

          Have you considered voting for one of the other political parties? If not, have you considered that maybe people like you are the root of the problem?

      • by blokkie (322983)

        there is also redhat's OpenShift.

      • This is a realm where 20 might not be better than 3. There are real advantages of scale here; if you suddenly need a LARGE ramp-up of facilities (your web site just got Slashdotted or mentioned on American Idol), a big provider can do it for you without breathing hard. A small company might not be able to provide the capacity you need on short notice.
  • I'd be far less trusting of Google when it comes to running every single analytic it knows of over my data. Amazon's got far less stake in regular data processing, they just want to know about shopping habits.

    • Re: (Score:2, Informative)

      by Anonymous Coward

      If you think that Google would try to (or even be contractually able to) mine data from their customers like this, you're either ridiculously paranoid or an anti-Google shill. Maybe both.

      • by Anonymous Coward on Sunday March 17, 2013 @01:28PM (#43197645)

        If you think Google (or any company) wouldn't try to monetize every bit of data that they get their hands on, you're ridiculously naive.

        • Yes. Therefore it makes all the more sense to avoid putting all your data into one basket. Google probably already has a lot of data on you, through their search engine and their countless other products. So in this case, I'd go with the non-Google offer.
        • by atom1c (2868995)

          every bit of data

          Of course they are, it is the underlying business objective supporting all of those Big Data research projects.

      • Re: (Score:1, Insightful)

        by Anonymous Coward

        You must be new to this world. Industrial espionage is a term that's not derived from science fiction. It's a real thing and it has been for a long time.

        Now imagine Google seeing a new emerging market they want to have a part of. This has happened before, think web mail and social media. Then imagine one of their future competitors is running their stuff on Googles network. Would you really think they could resist the temptation to peek into their stuff?

        Mind you, I'm not trying to bash Google here. Just abo

        • by Anonymous Coward on Sunday March 17, 2013 @02:10PM (#43197829)

          When it's a company paying for this service, Google will not be looking at their data.

          First, it's bad practice. Companies won't pay if they think their data is available to be read/hacked/distrod/etc.

          Second, if Google is smart, which I'm sure they are, the data won't even be in a format that google is able to decrypt. [wikipedia.org] They don't want to be knowingly storing potentially illegal things, so they will keep it in a format that they can't access without the client's private keys.

          • I don't think they will be liable for the illegal things customers store any more than they are liable for emails in GMaill of folks with criminal intentions.
            • One of the benefits of plausible deniability with regard to customer data, is that it gets you out of the 'responsibility loop' earlier. The longer you remain involved in any legal activity, the more cost you are accumulating in terms of legal representation, lost effort as your staff responds to supoenas, etc. This is something you want to minimize if you can't eliminate it completely.

              Let's consider a basic hypothetical situation in which 'Illegal Data' is discovered on a service provided by Hosting Comp

        • by Anonymous Coward

          Would you really think they could resist the temptation to peek into their stuff?

          Yes. Both because it would be a firing offense, and because Google is far too arrogant to believe that it can't do a better job of whatever it is anyway.

          (I work for Google.)

    • by Zadaz (950521)

      Amazon's got far less stake in regular data processing, they just want to know about shopping habits.

      Yeah. Right. Just like Google only cares about search results.

    • by Anonymous Coward

      From a previous Slashdot article:
      http://tech.slashdot.org/story/12/07/01/1523223/insights-into-google-compute-engine [slashdot.org]

      > Google offers at-rest encryption for all storage, whether it's local or attached over a network. 'Everything's automatically encrypted,' says Crandell, 'and it's encrypted outside the processing of the VM so there's no degradation of performance to get that feature.'"

  • Everything but CPU (Score:2, Insightful)

    by Anonymous Coward

    Benchmarks seemed to include everything but actual CPU (GFLOPS, Linpack, whatever) performance.

    I would hazard a guess that pure number-crunchers use less general-purpose farms (and probably farms of graphic coprocessors).

  • Limited preview (Score:5, Insightful)

    by hawguy (1600213) on Sunday March 17, 2013 @12:20PM (#43197275)

    You mean a cloud service in "limited preview" is much faster than a cloud server open to the public and heavily used?

    There much be some fancy engineering behind the scenes to make a lightly used service run faster than a heavily used one.

    I want to see the benchmarks after GCE is open to the public.

    • Re: (Score:2, Informative)

      by Anonymous Coward
      There much be proofreading before hitting submit.
    • I find it somewhat ironic that Microsoft will support Linux on its IaaS platform before Google will support Windows on theirs.

      • Re:Limited preview (Score:5, Insightful)

        by styrotech (136124) on Sunday March 17, 2013 @06:03PM (#43198957)

        I find it somewhat ironic that Microsoft will support Linux on its IaaS platform before Google will support Windows on theirs.

        Why is that ironic? The difficulty/pain for each of them to support the "other" OS isn't the same.

        Linux is easier for cloud providers. eg no license tracking, billing or activation type stuff (for most distros at least) to worry about, small Linux server instances require less resources than Windows, just a bunch of files to deploy - no installation processes, instance specific UUIDs etc

        Windows is harder (for everyone but MS) for the opposite reasons.

    • by swillden (191260)

      You mean a cloud service in "limited preview" is much faster than a cloud server open to the public and heavily used?

      There much be some fancy engineering behind the scenes to make a lightly used service run faster than a heavily used one.

      I want to see the benchmarks after GCE is open to the public.

      They'll be the same -- or maybe better, as the service continues to improve.

      GCE may be lightly used at present, but it has a massive competitor for compute resources: Google's own products. Even when GCE becomes very widely-used, it will still be small potatoes compared to Google's own compute load.

  • by BitZtream (692029) on Sunday March 17, 2013 @12:31PM (#43197337)

    Twice a day over 4 days ... 8 samples ... this is supposed to be useful in some way?

    You should be ashamed of yourself for presenting this data as if it has some sort of meaning at all, let alone a useful one.

    You're going to need a couple orders of magnitude more samples before you even start thinking about being any sort of useful metric.

  • by Anonymous Coward

    At least for the latency calculation, if they were comparable DC locations, the claimed latency number for google would be lower than the round trip speed of light time. The real answer is the google DCs being compared are adjacent whereas the amazon DCs are on opposite sides of the country.

  • Fast (Score:4, Insightful)

    by Anonymous Coward on Sunday March 17, 2013 @01:31PM (#43197661)

    It is fast because nobody is using it.

    • by swillden (191260)

      It is fast because nobody is using it.

      Google is. GCE runs on the same infrastructure that runs all of Google's services.

  • Support? (Score:5, Interesting)

    by gtirloni (1531285) on Sunday March 17, 2013 @06:26PM (#43199089)
    Has anyone had to interact with Google support for this? Is it anything like the other services?
  • How many users is each service supporting? I'd be willing to bet AWS is supporting 20x the user count GCE has, so if GCE is only 4x faster on some things, put an AWS-equivalent load on them and watch how quickly they fall over.

"Whoever undertakes to set himself up as a judge of Truth and Knowledge is shipwrecked by the laughter of the gods." -- Albert Einstein

Working...