Slashdot Mirror
New Prices For Google Apps Engine
First time accepted submitter oker writes "Yesterday Google announced Apps Engine would soon become a fully supported product. Unfortunately it also means a new pricing model, which is less appealing since free quotas have been lowered and you need to pay at least $9 per month if you go for the paid version."
It's Google App Engine. App, not Apps.
The mailing list has been awash in outrage and suprise as prices rise much higher than most can support. Although all knew the price increase was coming, the optimization done for the past models don't apply to the new pricing scheme, and the community is not happy about the quick change (2 weeks). http://groups.google.com/group/google-appengine/browse_thread/thread/a1b7c68db2243932/043cbc3b7c296d06?hl=en
If you're not going to tell us what it is, perhaps linking to something that does would be in order.
http://en.wikipedia.org/wiki/Google_App_Engine
Apps and Profiles/Google+?
Ya, probably not.
Mod me down with all of your hatred and your journey towards the dark side will be complete!
look reasonable to me.
I don't see how hating on this does anything other than discourage companies from offering public beta services.
It was never promised as free coffee for life or whatever the whiners claim they thought they were getting.
And it still has a free level!
The $9 minimum is instead of the $9 fee that they first announced, which effectively is a huge upgrade for the low end paid users.
Meh, at that cost ($9/mo?) - developers using it for non-android purposes will move away to a VPS. Giving the platform really just exclusivity for Android app makers for the publicity perks. The instance and bandwidth expenses are garbage compared to AWS.
The 9 dollars is just the base fee. If I'm reading this correctly the only thing you get for that $9/month is the ability to pay additional pay-as-you-go usage fees to get more resources; the monthly fee doesn't increase your inclusive resource allocation at all. If your application is actively used most of the time, you'll apparently end up paying quite a bit more...
If you spin-up an EC2 instance on AWS, with no additional features (like S3, SimpleDB, etc), it will cost you at least $15 a month. (They do have a "free" tier, but that's for the first instance, and not everyone qualifies).
The bigger question, of course, is usage. I need to sit down and really assess a normal app in terms of various resources used. I know that most of those services are dirt-cheap for AWS. (You have to consider that for a decent app, you wouldn't be using the "Micro" EC2, which is the $15 instance, but the Small instance, which start at about $75 a month or so)
The best thing about a boolean is even if you are wrong, you are only off by a bit.
I like the programing model behind app engine, wrote a few apps using it etc. However being locked to the platform you never had real control over sucked.
Today however if I want the same sort of infrastructure all I need to do is install tornado and redis on a ec2 or a rackspace instance and I have all the speed and bandwidth I need.
Got Code?
IIRC, they used to have 6.5 hours of CPU time free, now you get a full day. Doesn't help if you get Slashdotted, but it's still a significant difference.
Don't thank God, thank a doctor!
The price change is just outrageous.
Our simple logger application (4000 low latency requests (max 2 datastore ops per request) a day) will move from 9 USD per month to 288 USD per month (+3122%).
The main problems are:
1) AppEngine scheduler is not able to allocate instances economically (which was no problem, when instance time was free). "Always on" used 6 instances to accommodate the load "immense" load of 4000 request a day.
2) The datastore operations are really slow (a search by key name (in only 60.000 records can take over 100 ms))
3) Each instance can only process one requests at a time
4) The loader often fails with timeouts, so having instances preloaded seems to be essential, as soon a your application is moved from trivial to something useful
5) Frontend instances (with only 200 MB RAM) which used to cost 3.00 USD per month are now priced at 57.60 USD per month (+1820%)
Google's offer to only charge 50% of the instance time until end of October (when they think, that multi threading would be ready) will not solve the problems:
1) Instance time is much too expensive even when multi threading is available
2) The limited RAM size will become an issue with multi threading (maybe not for our logger application, but for most others)
1 Put frogs in pan of cold water 2 Increase heat slowly 3 ? 4 Profit
Korma: Good
At that price you'd think they would have added SSL domain support. Or at the very least had it as an option.
A kick in the teeth to Google+ http://www.readwriteweb.com/hack/2011/09/google-app-engine-pricing-ange.php
It used to be possible to burn through those 6.5 hours very quickly, and often for the same reason -- if you didn't have an instance constantly running, you'd burn CPU spinning one up if you got a request every few minutes, because the instance would spin down after a few minutes.
But if you ping it repeatedly, that's a significant amount of CPU lost just keeping it running.
An instance makes a lot more sense. If my app actually fits comfortably in the free tier, I'm probably not getting enough traffic for several simultaneous instances to be needed. Running one instance 24/7 for free makes sense, and is going to be a lot more responsive in that "a hit every few minutes" category of usage.
And while they might fire up another instance in response to traffic, that seems unlikely for a low-traffic app -- "1 request at a time" isn't a big deal when you consider that this concerns just the actual processing of the request -- other requests can pile up behind that, and you have a hard limit in how long a request can take anyway.
Don't thank God, thank a doctor!