• Print

Google App Engine Lets Your Web App Grow Up

app engine logo

Google released App Engine less than a year ago (Radar post). It was the first chance for external developers to use the power of Google’s servers. The powerful platform supported Python and was free (within limits). It now supports 45,000 apps and those apps get over 100 million page views per month day. Those pageviews were all free, but they had limits.

That’s going to change. After today developers can pay to have more storage, more bandwidth, more CPU time and send more email. The costs as of this morning are listed below with a comparison to the AWS equivalent cost.

• 10 cents per cpu core hour (AWS charges $.10/hr for a small, standard Linux instance and up to $1.20/hr for an XL, Hi-CPU Windows instance in EC2)

• $.10 per gigabyte transferred into AE (AWS charges $.10 for all data transferred into S3)

• $.12 per gigabyte transferred out of AE (AWS charges $.17 for the first 10 TB/month transferred out ofS3)

• $.15 per gigabyte stored per month (AWS charges $.15 for the first 50 TB/month stored onS3)

• .0001 dollars per email (AWS does not have an equivalent)

Without running a more advanced cost calculation it seems that App Engine is slightly cheaper for smaller web apps. Pete Koomen, an App Engine Product Manager, would not say if they would add tiered pricing. I can’t imagine it happening until after they add background processing for applications.

Developers will only pay after their app surpasses the free limits (and there are several AE apps that already have like buddypoke.com and mentalfloss.com). The existing free quotas will be reduced in 90 days. The new paid quotas do have unpublished limits, so if you need to support more than 500 requests/sec you’ll need to contact Google.

Developer will use their Google Checkout account to cover their apps costs. This means that the new features are restricted to the U.S. and U.K. only. Hopefully the Checkout team expands their scope soon. Google Checkout appears to be available almost everywhere, but doesn’t have a page that lists all of the countries. You can check the dropdown of the sign-up page to see if your country is supported and thus you can use the Google App Engine features.

This is a significant step for App Engine. By setting a theoretical limit above the dreams of many web apps they are now putting out the call for serious applications. They’ve already added Google App Engine Apps to Google Apps For Your Domain. I could envision AE becoming the backbone of an Enterprise Dev Marketplace.

AE still has many significant to-do items on their public Roadmap. Not the least of them is the support for more languages (Java is a good bet as it is used a lot internally) and to add support for background processes.

A huge concern with App Engine is platform lock-in. Google provides a lot of powerful, but non-standard APIs and features that make switching platforms difficult. Developers can extract themselves from App Engine via projects like AppDrop, but it is still risky to use their platform without an SLA. Without a guarantee Google could theoretically decide to raise prices unreasonably. Is it likely? No, but it is something that developers need to think about before committing to any platform.

Google I/O already has many sessions listed for App Engine developers. I’ll give a free pass to I/O to the person who suggests the best AE app in the comments. Vic Gundotra, who runs the App Engine team, will be at the Web 2.0 Expo SF in April.

Updated: Correct mistakes about the current pageviews and the availability of the new features.

tags: , , ,
  • http://bgraves.com Brian

    I’d like to see an app that automatically translates useful, common phrases (categorized or tagged for easy reference) in some common languages. There are websites out there that list how to say “Thank you” or “I love you” but I’ve never seen a more useful list with phrases like “Where is the nearest hotel?” or “Can you help me get a taxi?” or “Do you speak English/French/German?”, for example.

    I’ve thought a lot about this and haven’t seen anything like it online (at least not useful phrases in a well done website).

    Now I just need to scrape together the cash for my trip to San Francisco in May. Thanks O’Reilly!!

  • poko

    >This means that the new features are restricted to the U.S. and U.K. only.

    I think google checkout limitations are about merchants (“Google Checkout is available to U.S. and U.K. merchants”) not customers. So you still can buy stuff from a US based merchant via checkout with say a German credit card

  • http://www.littleshoot.org Adam Fisk

    Great the App Engine team has released this. The drop in free quotas is precipitous, I must say (10x less free bandwidth in and out, 8x less CPU time), but it’s understandable given the free quotas were quite generous.

    I’m not sure this is a “suggestion” for an app, but my own LittleShoot runs completely on App Engine, and we’re about to release the first BitTorrent browser plugin EVER (as in a true NPAPI/ActiveX browser plugin, none of these faux plugins in the form of FireFox extensions).

    That said, I’ll be at Google I/O anyway and would love to meet up if you’re heading there, Brady. Shoot me an e-mail any time.

  • http://rossnotes.com Ross M Karchner

    Here’s an app I’ve been thinking of– a “post-collapse skills inventory”, a neighborhood who-can-do-what registry of skills that might be necessary in a general (or localized) economic failure or lack of government services.

    Things like: first aid, growing produce or livestock, repairing engines or generators, use and access to firearms, foraging, etc.

    I’m not decided yet on how *sharing* that information would work (based on distance? town you live in? a social network), but members should get a printable weekly or monthly directory of whoever falls within that scope, and skills they offer. That way, even if ever google data fell off the internet, the data would still be out there.

  • http://rossnotes.com Ross M Karchner

    “ever google data” was supposed to be “every google data center”

  • Anurag

    Typo alert: Google App Engine ***Lets*** Your Web App Grow Up

  • http://jamtoday.org james

    Google Apps already has a growing user base from educational institutions, and with Google Apps Labs we already have a preview of App Engine applications hooking directly into Google Apps.

    With this in mind, App Engine could be a great place for educational applications, such as course management software, study guides, video lectures, and assessments.

  • http://jtarchie.com JT Archie

    idea: salsa venue finder

  • http://pphaneuf.livejournal.com/ Pierre Phaneuf

    For a small site, App Engine is a good deal cheaper: while the price of the CPU time might look equivalent at a glance, on EC2 you need to keep your instance up all the time, which is billable. So 30 days of usage comes down to $72, just for starters, whether you got any hits or not.

    If your app keeps a CPU load of 25% or less on average (on a single CPU machine, which is still pretty solid for a web app), it’s within the free quota.

    Of course, the other items are to be considered, but that makes a $72 baseline cost per month for Amazon EC2. I agree that it’s not the same kind of service, though, you can run all sorts of weird software on EC2 that you can’t with App Engine…

  • http://ini.hr Berislav Lopac

    One application I’ve thought of recently: a search engine for stock photo sites, such as iStockphoto or stock.xchng. It would display a number of photos from different sites, based on the keyword searched for.

  • bowerbird

    > I’d like to see an app that
    > automatically translates useful, common phrases
    > (categorized or tagged for easy reference)
    > in some common languages.

    crowd-sourced in a wiki, that idea would be very nice.

    -bowerbird

  • John Mark Mitchell

    I would say the best/most practical AE app that I have seen is http://downforeveryoneorjustme.com/

  • http://samj.net/ Sam Johnston

    Google App Engine is a perfect fit for grassroots campaigns like “Save the Netbooks” (http://www.savethenetbooks.com/) that need to be developed rapidly (under 24 hours in this case), to scale quickly when news of the campaign hits (e.g. Slashdotting) and then taper off to provide ongoing coverage indefinitely. All of this needs to be affordable for individuals/non-profits (ruling out alternatives like Amazon EC2) or ideally free, thus giving a voice to people who may never have had one before.

    The same could be said of any campaign (e.g. marketing) so a generic campaign support tool for everyone from individuals through small businesses, non-profits and large enterprises would be a “perfect match” for Google App Engine.

  • http://bgraves.com Brian

    @bowerbird – Yes, I think user contribution would be essential. That way, native speakers for a given language can modify a translation if it’s not quite from Google Translate…

  • http://www.joyent.com Bryan

    Brady,

    Thank you for summing up the recent GAE announcement. However, your post did keep me up last night (for a few minutes at least) because, like so many articles before, this perpetuates the myth that AWS delivers what GAE delivers and vice versa.

    EC2 is hosting – that’s it, that’s all. Yes, it’s by the hour and there is the concept of infinite, on-demand supply of infrastructure but, it still hosting. Even if you consider AWS storage products, AWS is hosting with additional products. It is nowhere near to a Platform as a Service (PaaS). It is so-much-so raw iron that until recently, third parties like Rightscale were the only ones supplying the management software to actually make operating it straightforward.

    GAE, on the other hand is a software development platform – a true PaaS. With GAE you get an application run-time environment on which you write code and once you deploy the code, it runs. GAE just happens to provide “hosting” as part of its service because that is what is needed to run an application. It auto-scales without any manual or third party intervention. And, GAE certainly does not provide general hosting like AWS.

    So, making a direct price comparison between the two is like comparing apples to oranges and perpetuates the myth that the two offering are comparable.

    With a PaaS you do not need third party software, network administrators or large consulting firms to manage deployment or ongoing operations. With AWS you do. Therefore, AWS is just more expensive to operate than software written on GAE’s PaaS (because you actually have to manage it).

    Hopefully, once GAE, and more importantly to me, the soon to be launched Joyent Smart Platform (F.K.A. http://reasonablysmart.com) evolve into open, language agnostic, fully distributed PaaS solutions, the value proposition becomes even more clear and this somewhat illogical comparison stops happening.

    I understand that in the software business, as in many others, some people make money by building something from the ground up and others just want to focus on making money by running a business.

    AWS is an offering suited for running legacy apps, replacing existing on-site infrastructure and for those who, for whatever reason, think they can build the entire stack better themselves (shameless plug: in all three cases, Joyent supports these endeavors better than AWS but, that is for a different discussion).

    A PaaS is suited to those who want to focus on differentiation, business building and strategy without having to worry about power generation.

    Once each of the value propositions is clearly understood the costs certainly will not be the same…

  • http://adjo.info Adam Jones

    I’d like to see a GAE app that people can embed in their blogs for displaying twitter messages.

  • http://diggerdesignlabs.com Steve Robillard

    I would love to see an app suite for testing and monitoring app development and deployments. It would automatically run tests (unit acceptace, performance etc.) against your app. Basically moving testing and monitoring from my local server (source control, build, continuous integration) and move this to a unified app in the cloud. It could even monitor availability and check for defacement etc. in live apps.

  • http://makingmoneyontheinternetwithgoogle.blogspot.com Brent

    I have had a little play with the Google apps but have not yet spent enough time to really have a solid opinion on the changes.

    As far as useful apps go I like the suggestions above re the local services directory and the camapign management tool.

  • Kunal

    This comment is to get free google I/O pass assuming it is still available. I would love to see a blackboard.com like application open to any teacher, school or university. The subjects can vary from 5th standard to java to appengine itself. The key components to be developed are grading, creating exams and open platform to extend! Also, the classes can be attended anytime in the future in which case the revenues goes back to the teacher or open sourced as per their settings. Where the lectures are posted using YouTube api, schedules are set up using calendar api and upcoming things can be shown using blog api. And GMail will become their defacto email system.
    We lack a good cloud service when it comes to the faster education.

  • Andreas K

    You have to add that you get a runtime environment that is not completely defined.

    Google does not even bother to define all limits that are builtin in the appengine, e.g. memory usage limits. Short googling suggests that the googlized JVM does lie about free memory too.

    As trivial these might sound, they are a problem, how can you plan your datastructures and algorithms when you do not when and why an OutOfMemoryError or MemoryError will happen?

    Furthermore, Google seems to offer no support with a reasonable response time. Problems that are common will probably get fixed quickly, but problems/bugs that are triggered only by your app or a small number of apps can lie around for quite some time.

    All this means that you exchange the cost of keeping an admin position for keeping a developer on call, should your app hit some limits that Google forgot to tell us about, for quick fixes and workarounds.

    Andreas

  • http://newsreports.org/ spenser

    As at October/2009, at least one large site hosted on the service was DDOS’d and it took 16 hours just for the site owner to be able to get past the frontline support to someone with enough juice to recognise the problem. This entailed a $400 payment for support.

    Then, it was reported that in fact, it was the storage facility that was crippled because the storage network is on the publicly accessible network interface.

    The site owner was asked for his cooperation by a google vp to suppress publication of the problem. A typical PR move. Public relations, not pagerank.

    There are a *lot* of warts they don’t want you to know about.

    The above incident was reported at theregister.co.uk