What big sites use Google App Engine? Have you tried it? Is it flexible enough to run big, serious projects on?
- It's easy to start with and fun to use. You shouldn't need more than one day to learn it.
- If you like to focus on coding and not worry about managing the server then you'll love it. But if you're a control freak and like to have access to low-level settings then you'll not.
- If you have prior experience with relational databases such as mysql, then you'll need an adjusting period to forget much of what you learned about databases and start thinking in new and different ways. Some things that are really easy in mysql, such as getting the number of records in a table or joining tables, become hard on app engine and require writing code.
- When it goes down, you'll feel helpless that you can't do anything to bring your site back up, but you'll also feel lucky that a Google employee is carrying that beeper and have to wake up and fix it.
- It scales, but it's not super fast. Due to its distributed nature, individual requests rarely finish in less than 50ms and usually take 100 to 150ms (that' my experience, yours will likely differ). But it can run a huge amount of requests at the same time.
- It solves the scaling problem, but it's not magic. You still need to optimize the heavily used parts of your code for scaling. But at least you do it with code, rather than having to fiddle with server and cache settings.
- It's a paradigm shift that will eventually eliminate most of database and system admin jobs if it catches it (I think it will).
- It's very different from Amazon's EC2. Amazon gives you servers that you have to manage, Google manages everything for you. With ec2 you pay per server hour, with App Engine you pay for resources you use: CPU Cycles, storage, bandwidth, ...etc.
- It's free to start with, but if you have heavy load it's a bit more expensive than hosting dedicated servers. I think it's worth it, though, because you'll be saving a lot of your own time.
- It's good for the majority of projects, but if your project has special not-too-common requirements, then it might not. Check first.
- It only supports python and Java right now.
If you're in Silicon Valley and if you need more in-depth information, you can find me at the next App Engine meetup at Google on the 6th.
Depends how well you want to learn it of course.
There's an open source version known as Open BlueDragon that runs on the GAE (and EC2) and has made the cloud their specialty http://www.OpenBlueDragon.org
Here's a link on the section of their wiki on the GAE: http://wiki.openbluedragon.org/wiki/index.php/Category:Googl...
And many other JVM languages, see this list:
http://groups.google.com/group/google-appengine-java/web/wil...
(which I notice is missing Clojure: http://fragmental.tw/2009/04/08/clojure-on-google-app-engine... )
It's been great for me. I love how quickly I can bootstrap new ideas, put up a prototype, and then just hand out the URL for to people to try.
A couple of recent apps:
http://create-a-password.appspot.com http://slimpoints.appspot.com http://mealminder.appspot.com
That last one was mostly a proof-of-concept in using Google's account and payment infrastructure. Really handy if you decide that you'll want your customers to pay you some day.
I guess the biggest hurdle is getting used to the Datastore, which is not really intuitive if all you're used to is SQL. Overall, a very positive experience for me.
Core apis I've had to patch: imaging (black lines on edges of pics), datastore (random timeout errors), inbound mail (improper body and attachment handling) and probably others that I've forgotten about.
There is also no portability at this point. I've seen one project that will run an app engine app on ec2 but I haven't used it.
I think it's a cool project but if you want the unhindered ability to build really good products I can't recommend it. Also don't mistake app engine for a hosting company. The experience for us has been 'you are lucky to be using our nifty cloud' and not 'thanks for your business, we're here to help'.
I've set up things on app engine and amazon as well. I found GAE just to be much, much easier and saved me a lot of time and money (yes I know EC2 is cheap but it's not cool to see your bill growing when you're just testing things).
I actually took the time to switch to EC2/S3 for a couple days because I was worried that GAE was not really serious about making it a productive place to run serious applications. Portability was a big factor too. But after a while I figured that all this worry was nonsensical. If I ever have to move things over, it must be a good sign and either someone will have built a tool for it or I just spend a week building a tool for it.
Also, re: SSL. You can actually do it right now on GAE but it involves a few tricky workarounds. That said, the GAE team just pushed a more straightforward feature for this to the top of the roadmap. Beforehand they had said they weren't going to work on it anytime soon, but presumably after so many people complained they decided to go ahead with it -- another good sign that at least they're responsive to developers' needs.
The times we are charged tend to be for burst traffic when something gets picked up on Digg or Reddit. Recently, one of our blog posts was featured on Digg, Reddit, Slashdot, etc. at the same time and Google dutifully served up a couple hundred thousand visits over the course of a few days. They charged us about a nickel! Granted everything is nicely cached so we were not doing much heavy processing.
It definitely depends on your web app. If you can cache it, App Engine is essentially free. If you are doing something CPU intensive for every visitor, it's going to be expensive.
http://code.google.com/appengine/docs/python/images/usingima...
Amazon Ec2/S3 works well as well and has the highest security bar of all the cloud offerings.
Rackspace is pretty good, but everyone seems very dis-satisfied with the S3 competitor, cloudfiles, so just use them for servers. Don't put anything super valuable there though, they know your root passwords (as does google, if you think about it, on GAE).
The users are on python. Apologies for the poor formatting/spelling, having to rush out the door.
URLFetch API - We’ve expanded the number of ports you can access with the URLFetch API. You can now access ports 80-90, 440-450, and 1024-65535.
Overall I feel GAE is very good to quickly validate your core ideas without having to worry about scalability at all. In SmsTweet I created the app in couple of days and as traffic grew, started optimizing little by little by using Tasks and Memcache. Yes there are a tonne of restrictions if you chose GAE, but there are decent workarounds for every one of them. You also have to learn stuff specific to GAE (datastore, tasks for example), but I guess they are trivial to pick up. And finally if your application does pick up traffic you can easily convert it to paid version and pay only for the extra traffic that you get. So a win win situation for you.
Both running Python + Django 1.1.1 (with Google App Engine Django Helper, http://code.google.com/p/google-app-engine-django/)
I don't know if they are "big", but you can see app engine in action there.
I recommend doing a fairly trivial project in your spare time to get the hang of it. Try out the following features before deciding:
- large data set related stuff like cursors, paging lots of records - transactions - GQL (look at the various limitations of it)
GAE gives you a set of solid conventions for writing a highly scalable app that you'd eventually have to deal with if you used a more traditional approach. Some can be a bit annoying at first but it's fairly obvious why each choice was made.
You might also check out jinja2, django, etc. But personally I love the simple webapp framework. It's clean, obvious, stays out of the way, and it's very fast.
Heroku! Basically the Ruby equivalent of GAE—you get a Postgres database instead of Bigtable, though (whether that's a good or bad thing is up to you.)
That config is basically only for development/prototypes, though (it has a 5MB database cap); their default tiny-ish plan is $15/month, which is quite decent for anything that runs Ruby, let alone on a grid.
Turning on all the knobs and levers on that page gave me "$5089/month", which is a slight bit terrorizing—but I can't imagine the kind of business that would grow to need that scale without becoming a multinational with its own IT.
This may warrant a separate question, but I've seen various posts here and there about running Ruby/Rails on GAE using JRuby. Does anyone have any experience as to just how realistic that is as a deployment option? I can imagine it being full of headaches, but I would love to hear from someone who has either tried it or is actually running it on a production app.
(1) On GAE you don't get to use some of the good parts of Clojure because you can't spawn any threads. So the rationale for Clojure has to be something else than agents and the Clojure concurrency model. I'd say use Clojure if you want functional programming and all of the Lispy goodness: code as data, macros...
For me Lispyness is a huge + so it was a no brainer.
(2) One of the things I really like about traditional Clojure web-development (using Ring or Compojure) is that if you set things up properly you can experiment with and patch a live system using the REPL. I haven't been able to set that up just yet on the GAE.
On your development server using the GAE jars, however, you can do it (see http://www.hackers-with-attitude.com/2009/08/intertactive-pr... for some guidance). There is nothing like being a C-c C-c away from updating your development server app :)
I can't emphasize enough taking the time to get Slime, your GAE development server, lein, and Clojure well-integrated. Your development cycle will make good ol' Java GAE development using the Eclipse plug-in (which I used to think was pretty sweet) feel really really clunky.
(See also http://elhumidor.blogspot.com/2009/04/clojure-on-google-appe... for some pointers on setting up Clojure on GAE)
(3) GAE's datastore works extremely well with Clojure because it's schema-less. You can store arbitrary hashmaps and structmaps in GAE with zero effort -- it's a match made in heaven. There are a couple of branches of appengine-clj (http://github.com/duelinmarkers/appengine-clj is the original, http://github.com/r0man/appengine-clj is a pretty good branch) that you can look at to get a feel what I mean.
(4) With GAE's transaction support and automatic index generation (it's good enough most of the time), you can dynamically create types (kinds), put them in the datastore, and query on them without having to do any of the traditional NoSQL stuff (thinking queries first, denormalizing ahead of time, etc.). It's very sweet. I've done some Clojure+Cassandra and Clojure+GAE is a lot more painless (as long as you can live with the other limitations of the GAE) -- although things are getting much easier in Cassandra world with Hector + automatic index generators.
(5) In terms of performance and robustness, nothing about Clojure's held me up. But since I haven't launched yet, that doesn't mean anything.
I don't know of any "high-traffic" Clojure+GAE apps out there -- if there are any, I'd love to get some numbers on costs (CPU, especially) to see if there is a significant difference with similar java projects.
(Actually, the guys behind https://the-deadline.appspot.com, see http://www.slideshare.net/smartrevolution/how-a-clojure-pet-..., just launched three days ago -- you may want to shoot them an email, they seem nice enough)
My guess is that this is a non-issue. Clojure holds up really well against Java and JVM languages like Scala in terms of performance (especially if you give type-hints), and I'd expect the same thing to be true on GAE.
(6) General Clojure limitation with web development: Clojure web frameworks are light. You don't get to start with anything as nice as Scala's Lift. You'll have to work and build your own tools -- the good thing is that Clojure is a blast and macros will, in the longer term, allow you to do much cooler stuff.
Compojure is a great light framework to start with. Also look at Enlive for templating.
On the flip-side Clojure immutable nature works very naturally with request / response.
(7) For the OP: before you get on GAE, spend some time reading about its limitations:
- exploding indexes (makes implementing text search very tough -- there is some irony in there I think)
- no threads (it kills me, I want to use pmap)
- not real message queues. GAE is a pull/request-based system, not a push system. It is, however, getting much better (tasks, comet on the coming features list). Still, if you are building a real-time system that needs constant background processing, I'd think twice right now.
- if GAE goes down, you're down until they fix things, and will be totally helpless.
- some of the GAE goodness/crack means you're locked in (e.g., index building -- to move away you'll have to build equivalent facilities for whatever DB you move to). Although you can protect yourself by being smart in how you abstract away the GAE.
(8) Can't think of anything else.
If there is interest, I have been thinking about writing some blog posts about my experience setting up a Clojure+GAE development environment, working with the data store, working the the GAE API, etc. Let me know here and I'll get a first post out of the way by week's end.
We can definitly recommend Google App Engine. The SDK is really cool and you don't have to care about managing servers. Scaling is easy. Its (almost) automatic. The datastore is a natural fit for Clojure datastructures, but you have to learn how to structure your data for a distributed key-value store: When you structure your data, you have to think about what has to be in the same transaction (entity-group) and at the same time you have to think about reducing data-dependencies to minimize contention. But when you have lots of data, this is what you have to do anyway. For me, it is much more fun than using an RDBMS. (Read this: http://code.google.com/intl/de-DE/appengine/docs/python/data...)
We did the whole system in Clojure, so you really can write large systems using Clojure/GAE. It works! (Although the UI is simple, TheDeadline is a complex piece of software e.g. containing a self-written rule-based expert-system, a production system, very similar to LISA http://lisa.sourceforge.net/.)
The functional paradigm means less code and makes it easier to build powerful abstractions. You don't have to write lots of boiler-plate code like in Java and Clojure is more concises and much more powerful than Python. (Stop! I don't want to start a programming language discussion now! :) ).
We will soon post some Clojure examples and some free software libraries for GAE.
If you are interested, just follow our blog H.W.A. (http://www.hackers-with-attitude.com) and/or try out http://the-deadline.appspot.com to get the "feeling". ;)
Anyway, are you being serious (i.e., in your experience is clj+gae noticeably slower than java+gae or python+gae?) or just funny? Since I've never done any serious profiling of Clojure on GAE, I'll bite and do some brief performance tests and report back.
I'd like to run some benchmarks for normal requests, where I doubt there'll be any real difference.
Thanks for the search.
I encourage you to write more about your Clojure/GAE work, and, if you do so, to reply here so that those of us who posted in this thread know when you've got something up. I guarantee you at least one reader. :)
I posted a screenshot of GAE quota usage on 430k reqs/day, might interest you: http://niryariv.wordpress.com/2009/11/27/google-app-engine-q...
It clocks in under 10kB compressed, is very fast, very accurate and I use it to test devices all the time.
Under current quotas, I'm able to serve ~200k users per day at no cost and am actively developing 2 more nano web tools.
The development environment is certainly restrictive, but the risk for web app prototyping is absolutely minimal.
I'll soon update that to use "Django non rel" though -- Django running on no-relational data models like Big Table:
http://bitbucket.org/wkornewald/django-nonrel-multidb/wiki/H...
I believe it is and am betting with my startup with hosted apps targeting SMBs with CRM (http://crm.ifreetools.com) and Online Database App Creator (http://creator.ifreetools.com).
If you are from RDBMS, at first you will have to become flexible to accommodate to the BigTable way of things - no group-by, joins, etc., - and find alternate ways of doing things. Once you make the switch, if will indeed be good enough for big or highly scalable projects.
FYI, there are few key things the GAE team is working on which can be found at http://code.google.com/appengine/docs/roadmap.html
Things like SSL for custom domains and background processing would make GAE even awesome to have your project running on.
Quick question, a bit off topic. Here: http://twitter.com/Classlet/status/3854900110 you say "We finally rescued our Twitter user name from a squatter!". How did you manage to do this?
As per some instructions on the web, we created a "classlet2" Twitter account. When logged in, there is a help form where you can open a ticket. It took a few weeks, but they eventually released the account to us.
They support a whole bunch of languages, but PHP is the only one worth using because their execution model is pure-CGI — no persistent processes. The pricing model is just like GAE: pre-pay into an account that is debited for the resources you actually use.
We're working on moving our application over to App Engine but waiting until they get SSL on your own domain before doing that. You can see the roadmap for it here http://code.google.com/appengine/docs/roadmap.html
I have some doubts about staying within the CPU quotas, since I have to take on a processing load in the tick-update loop of game time, but I think I can design it to be spread out somehow.
Basically you could forget about applications that needed to write game state to datastore in each step.
Even the simplest possible writes were already triggering CPU quota warnings [1].
I didn't touch GAE recently, so maybe they already relaxed it a bit, but when it started, it was usable practically only for read-heavy applications (for which Google strongly suggested to use memcached wherever possible).
Also moving larger amounts of data between your computer and GAE datastore was massive pain (e.g. backups or initial state load). You had to be very careful not to blow over your quotas, as such imports were doing a lot of writes.
---
[1] For example, in early days they counted CPU use in "megacycles":
* memcached based ajax request costed tens of megacycles
* datastore-read based ajax request costed hundreds of megacycles
* datastore-read-write ajax request costed thousands of megacycles