GAE pricing is not suited for smaller apps

[supercobra]

Good point about small apps I read on this group.

Most apps are not intended to be used by hundred of thousands or
millions of users so the point that Google makes, i.e. that
scalability is what we pay for on GAE, is not relevant for those
smaller apps. The vast majority of apps don't need load balancing and
therefore a simple AWS server, or Heroku Java setup can do a whole lot
of work and are 10x cheaper than GAE.

For example our typing race app www.typrx.com runs on App Engine. We
are growing the user base constantly but with the new pricing it is
really hurting us.

I think that Google App Engine won't be the preferred platform for
those smaller apps for a pricing like the one they just announced.

=======> Google: Fix this. <=======

-- super...@gmail.com
http://supercobrablogger.blogspot.com/

[Joe Bourne]

I'd have to disagree - AWS servers are an expensive option and dont
offer the "zero maintenance" approach that GAE provides.
If you rent an Amazon server, there's a whole bunch of stuff you
need to do to setup, manage and maintain your application, so if its
just a "small application" the cost overhead makes GAE a much better
option.

Also, if you dont need load balancing etc, AWS is the expensive
option. Once you price it out, a dedicated servers or VPS solution is
often cheaper as they very often come with free unlimited bandwidth.

GAE, (even at $9 per month) is a VERY cost effective, trouble free way
to host small web applications.

On Sep 3, 12:39 am, supercobra <superco...@gmail.com> wrote:
> Good point about small apps I read on this group.
>
> Most apps are not intended to be used by hundred of thousands or
> millions of users so the point that Google makes, i.e. that
> scalability is what we pay for on GAE, is not relevant for those
> smaller apps. The vast majority of apps don't need load balancing and
> therefore a simple AWS server, or Heroku Java setup can do a whole lot
> of work and are 10x cheaper than GAE.
>

> For example our typing race appwww.typrx.comruns on App Engine. We

> are growing the user base constantly but with the new pricing it is
> really hurting us.
>
> I think that Google App Engine won't be the preferred platform for
> those smaller apps for a pricing like the one they just announced.
>
> =======> Google: Fix this. <=======
>

> -- superco...@gmail.comhttp://supercobrablogger.blogspot.com/

[Tim Hoffman]

HI

I run a small business web site on appengine (since early 2009) www.fishandlily.com.au  its a dynamic website with a simple cms (also got some big apps on other instances) I was running "always on" which was  only $9 month, which protected me from the odd slow start up time on M/S and I didn't have to ping

the instance every few minutes.  For every other metric I am below the free thresholds even for the new billing model.

Under new billing if I left things the way they are it would go to about $90 per month, but to be honest I don't need 3 idle instances.

Just one instance is about all I need (min idle instances coming soon I believe).  Which if a single instance was permanently running takes me 

down to $30 which I find reasonable. Given the overall lack of headaches I have with appengine.

WIth a prefilled memcache, I can serve almost every request in  <60 ms (often down around 30ms) and less than 200ms if only starting up and I don't get a 

cache miss.  If I miss the cache and have to start the full stack it can be anywhere from 2sec -> 20sec depending on how M/S appengine is feeling ;-)

Cache miss on a warm on a warm instance is typically costing between 200 - 400ms unless appengine is having a bad day ;-)

I plan to move to HR which seems to provide more reliable startup times, and so I could possibly do away with a guaranteed single idle instance, 

as long as I run a request every so often and keep memcache warm.

So depending on what you are doing, I think appengine is eminently suitable for small apps.

This site is important too our business as it is how most people find our bricks and morter store, and we know if we don't get web traffic we don't 

get customers so even this tiny app (40-100 visitors a day on average, 100-250 pages views a day) is extremely important to my lively hood.

So it's not all doom and gloom here in appengine land. I feel there will be sweet spots for appengine under the new pricing scheme and some big misses,

Rgds

Tim

[Francois Masurel]

It looks like settings Min-Idle-Instance to 1 will in fact keep 2 instances (or more) running most of the time.

The scheduler will spawn new instance each time it receives new requests to keep this "idle" instance available.

Check these 2 related issues :

http://code.google.com/p/googleappengine/issues/detail?id=5765

http://code.google.com/p/googleappengine/issues/detail?id=5764

Francois

[Tim Hoffman]

Yep

I am sure they will have some teething issues ongoing with the scheduler. 

I have left things on automatic, and and dropped always on, and almost 90% of the time I have a single instance.

Rgds

T

[Tim Hoffman]

But definately the scheduler is currently screwy and needs some serious work.

T

[Francois MASUREL]

Yep just have disabled it this morning and I have only one single instance on most of the time.

François

--
You received this message because you are subscribed to the Google Groups "Google App Engine" group.
To view this discussion on the web visit https://groups.google.com/d/msg/google-appengine/-/pmkVLep9au0J.
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.

[stevep]

+1 re: scheduler.

The ironic thing to me is that when we got the lead-time announcement
in May, the wails of discontent were calmed by GAE engineers in these
boards citing how we should wait the see the affect Scheduler would
have. Google PR is obviously busy these days suggesting that the
current developer reaction is ill-founded because we all knew about
this in May. Perhaps ironic is not the right term to use here now that
we get to see the great Scheduler.

Having said all this, clearly Scheduler already has been vetted by
marketing for presentation to enterprise CIOs who will never
understand nor really care about its current inelegance. So...
"spitting into the wind..." I am afraid.

Regretfully,
stevep

[Anders]

It depends on what you mean by small. Many apps will still be able to run for free on GAE. But yeah, as soon as an app goes beyond the free quotas then the pricing may be way too high.

And companies developing larger applications may not want to use GAE because of the limited datastore functionality. So I'm not sure exactly what kind of customers Google is targeting with GAE. To speculate a bit: Yikes, maybe Google now has too few large customers and that they now on purpose have chosen to kill GAE, because they see that they will never have enough large customers whatever they do and whatever price model they use.

[Raymond C.]

Tim, I just curious why you think $30/month for hosing a site that has "(40-100 visitors a day on average, 100-250 pages views a day)" is reasonable?  I suppose you know with that money you can host over a thousand sites with that amount of traffic on a single machine elsewhere right?

Or just because you dont want to migrate out from GAE because you dont care to spend a thousand times more money?

[Tim Hoffman]

Hi

I don't want to migrate because I don't want to deal with os and complete application stack.

I don't want to have to wake and find something has fallen over.

Also I know I can reduce my cost below $30 a month, probably to half that, when I move to HRD.

I have had that site on appengine for over 2 years.  With virtually no problems. No disk full, no security patches etc....

(I have had to look after E10K's, I was IT manager for 300 seat call center and run mobile phone billing,) I know what it takes

to run this stuff.  I don't want to do any of this. I am in the process of leaving IT to do my own stuff.

Also $30 a month for piece of mind and no worries, I will take that any day, 

And finally we spend around $1000 dollars a month on Adwords, so $30 a month for no worries,

hosting,  why would I even think about moving.  My power bill for running a single 700 watt pump for a full year is $1400.

I won't pay for anything I see no value in.

T

[Raymond C.]

I think you are definitely the target users of GAE, having money to spend without caring the price. ("its cheap compared to the revenue i earn from elsewhere")

You raised some good points of GAE though, but my point is still "the price doesnt worth it".  You can host the same tiny site on heroku or AWS without costing you a buck.

[Tim Hoffman]

I can host on AWS without costing a buck for the first year with a micro instance.

Second year it will cost.

Also I need to keep ubuntu patched, my ebs volume clean (clean out logs).

etc....

AWS is not immune to outages (neither is appengine)

And heroku, well lets say I am a python person through and through, so heroku is not for me.

dotcloud is probably a better choice for me as an alternative to appengine.

Also I do care about what I pay for, however my time is worth far more.  I don't bother consulting for less than $100 per hour.

So like you say cost is all relative.

Rgds

Tim

[Tim Hoffman]

Hi Raymond, 

Lets look at some real numbers rather than broad statements.

In fact in the second year an AWS micro instance is charged at 0.02 per hour

So 

>>> 356 * (24 * 0.02)   

170.88

a reserved instance is

>>> 356 * (24 * 0.007)  + 56

115.808

vs appengine with a single instance around $30 per month so lets make it a round $400

But those charges for the AWS micro instance doesn't include bandwidth EBS volume networking etc...

admittedly they will be small, but the EBS volume of 8GB is going to be close to $1 per month current pricing

so we are up to around $182 and $127 per year. Not even half the cost of appengine, assuming I don't get my costs down there,

(But as I said before I am confident I can once I move to HR)

The $150 - $250 dollars a year I save on EC2 is not worth the extra hassle of maintaining a complete stack myself.

Lets talk about the stack on EC2.

Ubuntu

Ngnix

pyramid or bobo, or webapp2, ...  ( I personally am not keen on django)

an ORM (sqlobject, or storm)  and mysql or no ORM

and run ZODB, 

all patched, configured etc... by me and maintaned by me.

I can actually do all this in my sleep, been doing that sort of thing for many years, but I don't want to.

The only real upside I see is this solution gives me, is the ability to take it somewhere else, oh and at best a few 

hundred dollars saved. If I need to scale at all, I now need to move to a small instance, and that is priced about the 

same as appengine.

Rgds

Tim

[Vinuth Madinur]

Hi Tim,

I'm curious to know why the prices will be half after you move to HR.

Rgds

Tim

--
You received this message because you are subscribed to the Google Groups "Google App Engine" group.

To view this discussion on the web visit https://groups.google.com/d/msg/google-appengine/-/ojUoj5nBGnQJ.

[Francois Masurel]

Thanx for this interesting post, Tim.

I still think GAE is a viable solution for some low traffic web sites.

I managed a dedicated server for 3 years and it was really painful and very time consuming.

I'll try to stick to GAE for now even if my bill increases a bit (I was already paying for "Always on"  instances).

Francois 

[de Witte]

What do you mean with the limited database functionality. We are developing a large app with a lot of functions and so far no problems. Except that you have to reprogram your relational database mindset. (this sounds very geeky)

[Tim Hoffman]

Hi Gubbi

I feel I can halve them because feel a lot more comfortable if there are no instances hanging around and have them start on demand

on HR.  M/S startup has always proven to be problematic.

I have a test instance that I check every so often on HR and it has yet to fail to startup once or take 20+ seconds to start up.

Also moving to HR gives me the opportunity to use 2.7.  threaded requests means that I can serve more of the 30ms cached responses

from a single instance, which means it is less likely that more than a single instance will be needed most of the time.

My site really only gets traffic between 6:00am and 11:30pm (very locally focussed) so I could keep a single instance hot during that time, 

and assuming google sorts out the scheduler and lets me have no idle instances, I could let all the instances go from 11:00pm to 6:00am 

and have them started on demand.

So on that basis I firmly believe I can get down to less than $30 per month.

Rgds

Tim

[Tim Hoffman]

Why do we keep seeing these comments "Yikes, maybe Google now has too few large customers and that they now on purpose have chosen to kill GAE"

Where is the evidence that allows you to draw that conclusion?

or are you actually trying to start or perpetuate an internet meme and hope it comes true ?

[Anders]

Things like only one inequality match (such as less than or greater than) allowed in queries and no LIKE operator. The new full text search API may solve some of that.

[Anders]

I found this article: http://blog.labslice.com/2010/12/2010-cloud-computing-winner.html

It shows that GAE has a very tiny market share. Managers at Google may calculate that the market share for GAE will remain too low for them to invest much further in the product. I certainly hope my speculation is wrong, but it's good to plan for the worst.

[Tim Hoffman]

Didn't you read the going out of preview notice from google.

They garuntee to provide the service for 3 years. This gives organisations piece of mind.

And you are drawing the conclusion from that report which is very vague, that google plan to kill gae.  What have you been smoking ?

Its comparing job trends not platform take up.  And they list S3 (which isn't even remotely comparable), so how did they measure it?

They count the number of jobs that lists each of the names s3, ec2 azure and appengine.  So in 2010 there 

where less jobs requesting experience in appengine than the other three.

[Raymond C.]

Woo...Tim, your talk is getting annoying. "What have you been smoking ?"? I think it's a question you should ask yourself.

How about come back to the discussion after you have real usage on appengine? I mean, not hosting a site that get hundreds of pageview a DAY. Instead make a popular web app that functions for massive number of users. Why do you think you know the value of appengine so much when you are actually not paying much on it?

[Stephen]

On Sun, Sep 4, 2011 at 2:05 PM, Tim Hoffman <zute...@gmail.com> wrote:
>
> Didn't you read the going out of preview notice from google.
> They garuntee to provide the service for 3 years.

They have ALWAYS provided this guarantee. See section 10 of the Terms
of Service you agreed to:

http://code.google.com/appengine/terms.html

[Tim Hoffman]

Hi Raymond

This particular discussion was focused on small apps. Not big apps.  Note the subject "GAE pricing is not suited for smaller apps"

I have been and am involved with large apps. Thats a whole different discussion.  And I have repeatedly stated I am not talking about pricing at that scale in this thread.  I and I have not defended googles pricing on larger scale apps any where in these groups. 

The what are you smoking comment was directed at the conclusion drawn about job ads for cloud providers

as an indication that google would drop gae.  Such a conclusion I find just find untenable, based on the evidence in that report.

T

[Anders]

I wrote about the 3 years in another thread. My speculation is that this is a slow kill, because they can't make the kill obvious since media then would report a lot about how Google has failed with yet another product, this time in the important cloud computing space. And many existing customers would get outraged if they killed GAE too soon. So they let GAE die slowly by outrageous pricing and dwindling investment in the product, and instead they focus on cloud computing on higher abstraction levels such as Google Apps and Docs.

[Supercobra Thatbytes]

Personally I don't think it is a slow kill. I think cloud computing is a key business opportunity for Google and that their GAE product is second to none. The Google guys just got the pricing wrong. In fact, if they priced right and added few languages like PHP and Ruby and add a SQL database (SQL is in testing mode now) there would be not a lot of reasons to use AWS or Heroku.

[Anders]

Yes, could be that Google will keep investing in GAE, because it could have great future potential. Cloud computing WILL become huge, no doubt about that. Google making a strong investment in GAE would be really good. I love the simplicity and scalability of GAE. Maybe the new pricing is something some business managers at Google have come up with, above the heads of the developers, lol. :D Or the pricing may even turn out to be fair even though it looks a bit suspicious at face value.

[Raymond C.]

Yes I agree it definitely can has its future if doing it right.  Just the new pricing at this time is just forcing a number of existing heavy users out of the game and never look back.

[Paul]

I am certainly willing to pay more for all-in-one product. And it's
especially important early on, on smaller projects and start-ups. With
GAE you only need to worry about building your app. Costs of hiring
someone to maintain servers, worry about load balancer, patching etc
are huge for a smaller project. And sure, you can say that you can do
it by yourself - but at what cost? Programming less? Doing less
design? Slowing down the project? And if you plan to do any
advertisement or invest heavily into graphics then GAE costs will be a
small factor anyway.

I just wish that Google develops GAE faster and really invests in it.
I would like a true all-in-one service, withou a need to use tools
like PubNub. And I'd like other services integrated too [domain
hosting]. I was really happy to see them making micro-payment service.
Is it still US-only? I also hope that DataNucleus plugin gets a
revamp. Hmm... is there any wish-list where I could add those
things? :)

P.S. Google should really think about sponsoring some graduate
programs/courses and/or engineer's degrees.I may be making GAE app for
my degree and it's too bad cloud computing is not present at CS
courses.

[Tapir]

On Sep 5, 4:35 pm, Paul <pgronkiew...@gmail.com> wrote:
> I am certainly willing to pay more for all-in-one product. And it's
> especially important early on, on smaller projects and start-ups. With
> GAE you only need to worry about building your app. Costs of hiring
> someone to maintain servers, worry about load balancer, patching etc
> are huge for a smaller project. And sure, you can say that you can do
> it by yourself - but at what cost? Programming less? Doing less
> design? Slowing down the project? And if you plan to do any
> advertisement or invest heavily into graphics then GAE costs will be a
> small factor anyway.

hi, man, it is not a general all-in-one product, it is a general all-
in-one product
with many lock-in features.