I would be very interested to know how their entry level plan is free and yet generous.
I want it to be true, I want the numbers to work for them. I have a small site that does 70GB/month and this would somehow beat shared hosting (dreamhost in my case) at only $3/month for the extra 20GB. Yet this is even more generous than their phpfog offering.
One thing I did not notice was any mention of disk usage. Either way I'll at least test them. I've been wanting to try out PaaS for a while.
Because once you invest your time in their free plan, then you'll use them for every other site you have too. Heroku did this to me and so far it's worked out well for us both.
There's no mention of disk usage because it doesn't matter and has no effect on price. Pricing is based solely on RAM usage. The thing is, economies of scale in the infrastructure space have made this kind of pricing model extremely feasible on the PaaS side.
This could be quite a game changer. Other than the generous free allowance, the other big difference compared to Heroku is that it allows you to select the infrastructure you launch on, including three different global regions for AWS (Singapore, Ireland, Virginia).
By providing an alternative for Heroku's high prices and the "region-lock" (latency is always a big issue for potential customers), AppFog seems set to be a strong competitor. More competitors in these spaces are always welcome! =]
They might indeed do that. We'll see. Meanwhile, AppFog will be up and running on dozens of data centers by the end of 2012. Those currently available are just the beginning.
True, there are still glitches to be worked out. We just entered general availability a few days ago. I would encourage you to give it another shot. Errors of this sort tend to arise somewhat randomly.
Interesting, at amazon's regular prices that works out to about $12/month in services they are giving away (assuming 3 year reserved servers), but they'd have to cram 34 clients on to a server. But now you're splitting 26 EC2 compute units 34 ways ... thats less processor power then a Micro.
We are looking the gift horse in the mouth. It's free.
Yes, you will hit the limitations of the free plan very quickly indeed, at which point you should start paying.
The idea of these free plans is to let tire-kickers try out the system. They don't have to pay to test drive the UX and deployment mechanisms, which are the value-added features that sell a platform like Appfog. They don't have to pay for low-traffic prototypes. They can launch a whole set of minimum-viable sites on free hardware and then upscale one within a few minutes of any traffic spike (in theory, anyway). And it empowers marketing maneuvers like getting an entire programming class of 25 people using your platform for a one-day class.
(I'm not sure I love the above business plan, to say the least, but that's the idea, and thanks to Heroku it has become the standard.)
I think that that's a good description overall, although I would add that the free plan allows deployment to all currently available IaaS and hence to do things like run analytics on the different infras to see what works best. But I would push back a little bit by saying that you can run apps getting some pretty legit traffic for free. It needn't be just "Hello world" or sandbox apps.
Free plan is too good to be true, it reminded me of Google Apps, their free plan started with unlimited users and they ended up with only 10.
let's see how long they can afford this generosity,
The sister concern: phpfog charges [5$ a month for custom domains](http://appfog.com/products/phpfog/pricing/) on the same tier. I would guess they'll have to eventually pull a switch on that.
I once tried to setup Symfony2 framework in PHPFog which is sister concern of AppFog. It did not went nicely. As Symfony2 generates classes in cache folder adding permission to that folder was tricky. Also have to clear the cache to push new changes. Dotcloud is better on this case. It gives shell access to both web server and db server.
After migrating a Java application that ran on EC2 to AppFog over the weekend, I wouldn't recommend running mission-critical apps on their infrastructure yet. A few things that I ran into
- After EC2 East was slow like molasses, I switched over to EC2 EU which was still speedy
- That caused their CLI tool to fall on its nose when trying to tunnel to the database (right now the bridge always gets installed in EC2 East). Fixed with a simple patch in the Ruby source
- Looks like the backend connector in nginx won't connect to your app if you have Basic-Auth on your root index
- Later MySQL became unavailable with "Host '10.0.47.186' is blocked because of many connection errors; unblock with 'mysqladmin flush-hosts'". No connection limiter between different apps?
I think AppFog is offering a great value should they be able to iron out all the kinks. I'm crossing my fingers that this will happen quickly.
I emailed them about this and they suggested placing my data in RDS or hosting it outside of AppFog in some other way and running only my app servers with their service.
dotCloud (http://dotcloud.com) is a paas that is natively multi-az. You can use that feature to reliably deploy ruby , php, nodejs, python, mysql, mongo, redis etc. It also has real support and a sustainable business model.
AppFog has "real" support (including live chat with all plans, including free plans). And the business model will prove to be quite sustainable (and bundled with a readily comprehensible pricing model).
Heroku Java deployment is "unusual" and there are many Java developers out there that prefer WAR/EAR that complies with JEE5 (legacy purpose) or JEE6. If you can achieve that, get ready for massive users, I know there are people who don't like GAE model for various reasons.
Basically: make it like Jelastic and don't limit it like CloudFoundry (no EJB 3.1, etc etc).
JavaEE deployment model is preferable over some "weird" setups.
I'm not much of a Java person myself, but our dev team has told me that the WAR + Tomcat route tends to work best on AppFog, at least at this point in time. This is also our recommended way of running other JVM-based languages.
Your dev team is spot on. The Java market does not have a solid Cloud Platform so far because most vendors go their own way to support Java (Heroku and CloudFoundry are the example of going their own way).
The only Java-friendly PaaS so far is Jelastic simply because they support GlassFish and Tomcat utilizing standard deployment model. But Jelastic does not provide the infrastructure (data-center or AWS, whichever), they partner with hosting providers, which I think is not a great move.
If AppFog can come up with a more standard/sane Java PaaS model out there, you guys can win that particular market.
If I may suggest more:
1) Tomcat 7 instead of the older 5 or 6.
The JavaEE landscape has changed significantly from J2EE 4 to JEE 5 to JEE 6 in which each major version moves toward _way_ less code and reduce XML/settings/configurations to a very minimum.
2) TomEE (if possible, or as "add-on"/next-level offerings)
Tomcat 7 implements a subset of JavaEE 6 (known as the "web-profile": Servlet[controller], JSP[template], JSF[view/controller], Dependency Injection, JPA[orm])
Enterprises who are still in love with EJB 3.x need TomEE.
PostgreSQL does not seem to be available, despite their ad copy. Googling reveals that it's been in beta and they seem to be expecting to roll it out soon.
Rather than saying "etc." which makes it look like there are additional databases you currently support but are not listed, why not change it to "and more on the way" which clearly communicates what you mean?
Try going into the console and seeing if the app is running anyway. Sometimes 500 errors spring up superfluously, even when things run fine. A small hiccup that we're working on aggressively.
We worked hard on Sinatra support this past weekend. Give it another shot. I've deployed multiple test Sinatra apps painlessly today, which was admittedly far less painless last week.
Can someone explain the concept or value of multiple instances on this? If an app is getting hit pretty hard can you just turn on another instance, albeit with less ram? Do you have to write your code in such a way to support it being run on multiple instances? Are there certain types of apps that are CPU intensive and RAM light, making the many instance/not much ram trade off a good idea?
I couldn't totally understand pricing model. You can get only twice of free plan when you agree to pay 100$ a month. Hetzner still seems logical, if you have enough passion to learn little bit server administration.
http://www.hetzner.de/en/hosting/produkte_rootserver/ex5
Hetzner is simply not a platform-as-a-service. It might be a good deal for a server, but that's not what you're paying for with a PaaS like AppFog. What you're paying for is application lifecycle management, from deployment to scaling to cloning to re-deployment on a different infrastructure, to adding services (like Mongo, MySQL, etc.) at the click of a button.
Serverraum alleine = etwas ganz ganz ganz anderes.
Agreed. As I write if you have will to learn Linux server management, you can easily achive what it offers. I can use PaaS if the price as low as not worth to do server administration. With same price point, I can have a real hardware with 12x more RAM and 1.5 TB disk cap and I can also add a skill to myself which is a good thing for long-term. Nowadays adding a service to your infrastructure is just an apt-get command. Deployment is easy as a git push. I totally understand who don't want to deal with terminals, config files etc. But as I said for me it is far from useful.
No integrated git-push deploy (I know you can use a gem with an extra instance to do that, but still). No SSH into your app. No custom app. Extra runtimes are there, but old (Erlang, for example), and no idea how to use them.
The AppFog command line tool is very straightforward and even somewhat Git-like: http://docs.appfog.com/getting-started/af-cli. If the "af" gem is installed, "af push" is the equivalent to a "git push." You're right that it's not the same thing, but it's pretty handy.
Extra runtimes (newer versions of Erlang, plus Perl, Go, etc.) will become available over time. They're high priorities for our dev team, as are Redis, Memcached, etc.
Regarding git vs af, if I could use the git already installed in my computer, and by simply adding my SSH key into AppFog system, I could avoid using `af` gem and Ruby altogether. Otherwise, I would have to install Ruby just for this purpose. But this is just personal taste.
Other PaaS such as Heroku or OpenShift allows you to set up a custom app. For example, you can create your own app type in Heroku's cedar stack with buildpacks, or the DIY app in OpenShift with supports for hooks.
Erlang probably isn't on the top of the priorities list, but I'm looking forwards to it.
No where on this page does it REALLY explain what this product actually does. Marketing/copy fail. Even the about page you have to click around to find is more about their funding than what I gain from this product.
Been using a lot of Dart lately, so I'm wondering if I could get a Dart VM running on AppFog. Since it's based on Java, there shouldn't be a problem, right?
I want it to be true, I want the numbers to work for them. I have a small site that does 70GB/month and this would somehow beat shared hosting (dreamhost in my case) at only $3/month for the extra 20GB. Yet this is even more generous than their phpfog offering.
One thing I did not notice was any mention of disk usage. Either way I'll at least test them. I've been wanting to try out PaaS for a while.