New App Engine Pricing... $1411/yr without processing a single request??

[Rori Stumpf]

Right now I am paying $9/month for an always on app... the intent was
to develop an app, test it with a small group of beta users, and then
eventually scale up - whilst being able to project (and predict) costs
in line with usage.

The idea is to be able to experiment with PAAS at low cost, and then
also pay for the service at the same (or better) ratio to increased
resource usage. But this new pricing model blows it all up and
entirely changes the way I see App Engine for experimental (and
production) projects:

In my case it seems the cost will balloon to $1.96 x 2 x 30 = $117.60/
month... and that's the price even if ZERO requests are processed by
the app.

This does not make sense at all.

I really liked App Engine, but now the annual cost has ballooned to
$1411 before processing a single request.
Does Google only want big apps on their systems? I understand that App
Engine needs to be profitable, but this seems like a move to kill the
project.

(below is, apparently, pricing at a 50% reduced rate through November)
Resource Used Free Billable Charge Frontend Instance Hours:
$0.04/Hour 72.77 24.00 48.77 $1.96

[Barry Hunter]

One thing to realise about the before and after 'prediction', it just
shows what the cost would be at the current settings.

With a few simple tweaks to the settings for the application, you can
probably reduce the price substantially.

http://code.google.com/appengine/articles/managing-resources.html

(eg, 'always on' means you have 3 instances always running. With the
settings you can bring it down so that there are only instances being
spun up to serve requests. Can make it so there is still an instance
ready to serve request - ie one instance 'always on' - which would be
your free instance. Then when you have spikes of traffic, you would
pay for extra instances. These might well fit in the $9 month minimum
spend anyway. )

Another thing, if you really do want to keep 3 instances 'around' can
buy reserved instance hours, so the cost will only be 2.5c an instance
hour.

And you get a $50 credit, now, which will help 'tie you over' until
multithreading is generally available. which should mean even less
instances.

> --
> You received this message because you are subscribed to the Google Groups "Google App Engine" group.
> To post to this group, send email to google-a...@googlegroups.com.
> To unsubscribe from this group, send email to google-appengi...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.
>
>

[Rori Stumpf]

That's how I *thought* it would work too, by when trying to change
"Max Idle Instances" to 1 or 2, the setting does not stick - it is
flipped back to "automatic" after clicking the save button. So "always
on" would cost at least $1411yr. I probably have to switch to a more
responsive platform.

[Gregory D'alesandre]

Hi Rori,

We haven't be able to reproduce the problem, would you be willing to go to this issue to fill in your browser information and the user you were using to try to do this: http://code.google.com/p/googleappengine/issues/detail?id=5751

Thanks!

Greg

[Rori Stumpf]

Hi Greg,

I'm somewhat pleased to hear it's a bug :-)
I've posted the info you requested, let me know if you need anything
else.

Thanks,
Rori

[Gregory D'alesandre]

I think we've tracked the bug down.  We are currently treating Always-On as if it is 3 min idle instances.  When you try to set Max Idle instances to less than 3 (which wouldn't make much sense if you had known you had set min idle instances to 3), the Admin Console doesn't let you and (unfortunately) doesn't give a good error as to why.  So, you can set it to 3, but as long as Always-On is still on it won't let you set it to less than that.

We'll see if we can figure out a way around this...

Sorry about that!

Greg

[stevep]

Greg,

Can you formulate a reason why the minimum of 3 instances should be
left on given The Scheduler's new role?

I had though "min 3" an artifact of the old $9/mn plan, and that
anyone with a low-volume site looking for minimum pricing would set
max idle to 1, and then vary latency slider according to how much he/
she wants users to be affected by new spinups should a burst of
traffic hit the site.

Thanks,
stevep

[Rori Stumpf]

I'm *really* hoping you can figure something out. I really don't want
to drop GAE for many reasons, but I won't have much choice if I'm
forced to pay for 3 "always on" instances that are not needed. If my
new app gets enough users then it will make complete sense from a cost
standpoint, not to mention all the other advantages of the platform.
But what about my costs while I am building the customer base...
should I find some other solution that will smooth out the cost curve?
(and most likely end up staying there).

Of course, if the spin-up's were more predictable, then I could live
with that and would not need "always on". But my current app needs to
(preferably) respond within 2 secs. And 1 instance would be just fine
for that.

Also, as I mentioned before, I think many GAE users (including myself)
would prefer affordable pricing for low volume apps, as that
encourages experimentation and proof-of-concept testing on the
platform. In turn, that should translate into much broader adoption of
the GAE platform.

Doesn't Google want to lock startups into GAE at the embryonic (i.e.
no money, just an idea) stage? ;-)

[Jeff Deskins]

Rori,

I ran into this same problem earlier - where I paid for always on and couldn't drop the min idle instances below 3.  The new pricing showed me paying for those extra instances that weren't doing anything.  

Since my app is starting to get a steadier flow of traffic now, I disabled always-on feature and created a cron job to basically ping the app every couple minutes to make sure the app is still active and memcache is populated, etc.  This allowed me to drop down to 1 min idle instance with no issues and am able to see a more realistic picture of the new pricing charges for my app.

Jeff