GAE vs Heroku

GAE (Google App Engine)

The migration is almost impossible so you're married with google and its future decisions
All applications have a free amount of usage (reset daily) and Paid applications can use more at the prices below. Paid application: $9/app/month
Pricing page: Google App Engine Pricing

Heroku

You could migrate, so if for some reason you decide to stop working with heroku you could change of supplier
Is more easy to find developers with experience in one of the Open Source frameworks supported by Heroku than find someone with experience in GAE
Pricing page: Heroku pricing

Web

• GAE: On-demand Frontend Instances $0.08 / hour
• Heroku: Web Dynos $0.05/hour.

Database

• GAE: High Replication Datastore $0.24/G/month

• Heroku:

• Basic 10M rows, $9
• Crane 1TB database, 400 MB Cache, $50
• Kappa 1TB database, 800 MB Cache, $100
• more cache, more money...

In database the pricing by GB of GAE is more flexible and more cheaply at the beginning than heroku but things change when you got to 170 GB

Datastore API

• GAE
• $0.10/100k write ops
• $0.07/100k read ops
• $0.01/100k small ops
• Heroku: Free

File Storage

• Heroku: 
• S3 Standard Storage: $0.095 per GB
• S3 pricing
• GAE:
• Space used for Blobstore values contributes to the Stored Data (billable) quota. Blob info entities in the datastore count towards datastore-related limits.
• High Replication Datastore     1G/$0.24
• Blobstore API     $0.13/G

S3 is a lot more cheap than GAE

Outgoing Bandwidth

• GAE: $0.12/G
• Heroku: for S3 files
• Up to 10 TB/month     $0.12 per GB
• Request Pricing: PUT, COPY, POST, or LIST Requests     $0.01 per  1,000 requests

Email

• GAE: $0.01 / 100 recipients
• Heroku: Mandrill by MailChimp Emails per month 40,000 for 10 USD

Is more cheap to send mail from GAE

Background jobs

• Heroku: Worker Dynos $0.05/hour
• GAE:
• Task Queue Stored Task Bytes: the bytes consumed by tasks the application has enqueued that are not yet executed. This quota is counted as part of Stored Data (billable).
• High Replication Datastore 1G/$0.24

Summary

GAE have a more flexible pricing but its also more complex and with more items. A typical user of GAE is in one of this schemes:

1. Free
2. $9 USD + On-demand Frontend Instances $0.08/hour + High Replication Datastore $0.24/G/month + etc...

Heroku have a more simple pricing, but isn't that flexible in database prices. A typical user of Heroku is in one of this schemes:

1. Free Dyno + Free DB 10K rows + S3 free Usage Tier
2. Dynos $0.05/hour + $9 USD monthly Basic 10M rows + S3
3. Dynos $0.05/hour + $50 USD monthly 1TB database with 400 MB Cache + S3

Looking to this data my educated guess is that the more cheap option for a common web site with storage of database and files:

• with less than 120 GB: GAE
• with more than 120 GB: Heroku

A more in depth analysis

Compare PAAS is pretty hard, in a new post I will like to make a more detailed comparison with

• How much cache gives GAE to the DB
• Capacity of the web instances
• Administrative interfaces

All the comments are welcome, this is a pretty hard issue and I'm just a rocky on it