Yeah, I have mixed feelings about that. On one hand, I like free. If you are maxing out your free quota (which I'm not even close to), the CPU adjustment means what was free will cost you (46.30 - 6.5 cpu/hours) * $0.10 = $3.98/day or about $120/mo. The bandwidth adjustment drops 9GB/day at 0.10/GB in, 0.12/GB out -- that's $1.98/day or about $60/mo. There's also the issue of how much CPU burn a datastore write takes:
http://groups.google.com/group/google-appengine/browse_frm/t...
On the other hand, I want a well-supported platform from Google and if reducing the free stuff means they have a good business model, can hire more engineers, etc, then that's a good thing.
To be fair, at 300Gb/month of free stuff they would have killed most web hosting companies out there.
Most web hosting companies to not depend on extraordinarily price-conscious highly technical people for their bread and butter.
Could you have gotten Wordpress working on Google without having an engineering degree? Could you have set up a simple brochureware site without an engineering degree? These are common customer requirements among the cheap, mass market web hosting companies.
What's interesting to me is that there are still hard limits for billing-enabled apps.
There's the respectable ones: You get the full use of a 100 megabit pipe, 72 CPUs, and a 500 req/s proxy; plus you can send millions of emails a day.
On the other hand, the limits on datastore and memcache access are out of proportion to the bandwidth and request limits, especially when compared to the free limits. With a paid app you're unlikely to be able to ever hit the Bandwidth / CPU / Request limits because you'll be starved for data, whereas with a free app the opposite is true.
They prolly watched whatever real apps they had and determined appropriate balance based on that data.
But then some things look very odd. Like: "data sent to datastore API" capped at 153Mb/min, which is 2.5Mb/sec. Uhm. I don't think this qualifies as a distributed high-scale system, if that's the best they can do...
I don't think it's a straightforward comparison. App Engine to me is similar to Heroku or EC2 + RightScale. It'll be interesting to see Heroku pricing since they've got to charge more than raw AWS pricing. The cost of raw resources from AWS and Google are similar -- depends on what you are using on the AWS-side for data storage.
This can happen anywhere, with any hosting solution as long as you don't have physical access to you servers. AWS, Linode, Slicehost...they all can go down.
You can have physical access to your servers and still watch them go down. Your upstream bandwidth provider had a technician trip on a power switch. Your city suffers a Katrina-level disturbance in normal operating conditions. etc, etc Its all a matter of what risks you tolerate.
Anecdote: Google Search has had more downtime in the last 400 days than my main Slicehost slice. This demonstrates to me two things: a) Slicehost is awesome and b) you are the weakest link, whether "you" are Google's whole engineering team or one guy coding from next to a rice paddy.
Personally, anything I code against these APIs, I make as encapsulated/swappable as possible. Downtime aside, I am worried about being cancelled or if someone goes out of business or if someone starts offering a piece of the puzzle that will save me a signficant amount of money. Got to be able to adapt as fast as possible.
Fortunately, Linux lock-in is not something that keeps me up at night :-). This is the beauty of IaaS (VMs) vs PaaS (framework X).
There's always a degree of lock-in to any setup. That said, you can minimize lock-in by going with Django as much as you can and providing an abstraction layer to whatever APIs you use. Open-source key-value stores are popping up everywhere, so non-Google-hosted App Engine frameworks will probably appear in the next year. (AppDrop provided a fast check on porting App Engine SDK to EC2.)
Haven't heard of AppDrop - thanks! I can't wait for Ubuntu 9.10 and other projects that make building your own cloud easier. That'll bring a much need dose of competition to the market.
Question: is anyone using this as hosting instead of building a webapp. Is that against the terms? I've thought about just putting up a relatively simple slightly static django page and using google app engine as hosting.
There are some CMS AppEngine apps out there [1] so if you want to host a simple brochureware site, it's definitely feasible. It's still not "one click" simple though, namely you need to know how to checkout and deploy the app. I'm wondering why nobody writes something to give simple one click access to such a CMS. Hmm that gives me an idea... One click access to any AppEngine app given its source repository. That would be nice.
There seems to be an inconsistency in the cited storage pricing. The blog post says $0.15/GB/mo, but the docs and my app dashboard (which are most likely to be correct) say $0.005/GB/mo, which is a huge drop if true.
This is really, really significant. The fact that you can go above quota, and pay for the privilege, makes me far more likely to consider using this product. I might even get around to learning Python...
There are other good reasons to learn python, namely the thrilling, liberating ability to make useful programs quickly and without having to write and maintain a lot of pointless glue code (looking at you here, Java, C++ and co).
GAE was that which tipped me over to learn python and I am very glad now.
