The *real* cost of the new billing prices

[James Briant]

The real cost: My professional reputation

I've spend the last year or more advocating GAE in the San Diego startup community. I have pointed out that Heroku is expensive compared to GAE. I have pointed out that Heroku doesn't scale automatically. In return, my Ruby friends pointed out that GAE is a bunch of hoop jumping and vendor lock in.

Now I have to go the next Ruby meeting and be The App Engine Guy. The Heroku team was there last month. I have to show up at Tech Coffee Morning and be The App Engine Guy. I can't wait for the next Startup Weekend. I cant wait to talk to my Angel Investor friend who only two weeks ago I advised to use GAE over Heroku. Awesome.

I am feeling very naive, very foolish, and very angry.

[Raymond C.]

You are not alone.  I suppose EVERY AppEngine guys should be regretting what they have said and done in the past 3 years.

[Tim Hoffman]

Absolutely not.

I don't regret a single thing I have said about appengine.  I always explain the risks too. I have been here a long time, and been pretty vocal about shortcomings.

T

[Raymond C.]

Glad there is someone not.

Tim, would you mind sharing your data on the price change? I mean the numbers that makes you still think GAE is a good platform for you to stay.

[Tim Hoffman]

Hi Raymond

Here you go, posted about it earlier today.

This is at the low end of the equations and numbers talkd about around here.  But appengine still works for me. 

https://groups.google.com/d/msg/google-appengine/0tKkjM5xlc4/1bmSJx5huvwJ

Rgds

T

[Tim Hoffman]

Please note, I am not an apologist for the new billing regime, I think there have been some major misteps with it, 

especially with regards to clarity of information , lack of tools to really gauge the impact of your own chnages.

Billing stats 3 days old is pretty useless if you trying to reduce you costs.

But I see value in the service moving forward.  There will be certain app sizes and profiles and business models that just wont

suit appengine moving forward. 

But there has been way too much flung around about how one admin can make a super scalable AWS instance for less than google.

I just don't believe it is feasible for most people/organisations.  I have been doing IT for 26 years.  My first web app was perl/c front end to a

CICS/Cobol mainframe app communicating over LU6.2 on token ring from a Sun box in 1994.  So I have seen it all, and done most of it.

Rgds

T

[Bay]

In your post you claim that everything will be okay (even though you jumped from 9USD to 30USD pr month = 420 USD pr year). The basis for this is that you say that you do not really need the 3 instances running - only 1. Thats all mighty fine - but the problem is that you can't get it down to only 1 instance. Try and you will find out that the scheduler is rigged to always keeping around 2-3 instances running.

But again - if you're okay with 420 USD pr. year for something that doesn't really need the scaling (my estimate from your quota numbers) and can be served at any other site for 60 USD, then I guess the new costs really arent an issue for you.

[Raymond C.]

Thanks so much for your reply.  Congratulates you still find yourself fit in the new pricing model.  Unfortunately I dont find it myself.  The money I will be paying for under the new price allows me to rent several powerful servers on rackspace or EC2, just to serve 30-60 requests per second. (The high cost is due to high amount of small requests plus large amount of database writes)  

I am ok to build my own stack since its just one off and it also gives a lot of flexibility (like I dont need to rely on UrbanAirship for iPhone push notification anymore, thats another cost).

Auto scale is always a good thing, but not with this ridiculous amount of money.  There are so many free or low cost replacement on the other cloud platform that do the same thing.

[Tim Hoffman]

Hi

Actually I believe I can get my price down below $30 a month, keeping a single instance running will cost around $30.

I can get down to single instances no problems.  Sure the scheduler isn't helping.

Total number of instances	Average QPS*	Average Latency*	Average Memory
1 total	0.017	61.0 ms	49.9 MBytes

Instances 

QPS*	Latency*	Requests	Errors	Age	Memory	Availability
0.017	61.0 ms	286	0	1:52:20	49.9 MBytes	Dynamic

* QPS and latency values are an average over the last minute.

So I predicate (need better tools or I wait for 3 days) I can get it well under $30, and for me the lack of OS and application stack hassle

is a big plus.  My time is worth a great deal more, so if I don't have to worry about all that other stuff, then I am very happy.

See ya

T

[Nick Rudnik]

If you can only keep it down to one instance, it will be free. You get 24 instance hours per day as part of the free quota. But god help you if you think you can keep it to one. The scheduler can and will at any time fire up additional instances seemingly without reason.

[Joshua Smith]

It's not "without reason."

The scheduler behaves quite predictably.  Just look at the logs and find the requests that start up new instances.  You'll see that they are bunched up with other requests.  And there is a good chance they are coming from task queues.

On Sep 3, 2011, at 12:05 PM, Nick Rudnik wrote:

If you can only keep it down to one instance, it will be free. You get 24 instance hours per day as part of the free quota. But god help you if you think you can keep it to one. The scheduler can and will at any time fire up additional instances seemingly without reason.

--
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/-/ZriV8oWiNcwJ.
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.