6 comments

[ 2.9 ms ] story [ 23.7 ms ] thread
The email sent to existing customers mentions price drops for certain plans (text below). With the few plans I glanced at for 2.0, it doesn't look so much like a price drop as it does a storage capacity and max connections drop.

Production plans used to allow 1TB of storage and 500 connections. Now the smallest production plan has dropped to 64GB of storage and 60 connections.

I know Heroku is trying to run a business, and that computing resources do cost money -- but that copy seems a bit misleading to me.

Customer email blast:

""" As someone who is already using a Heroku Postgres production database we’re excited to ensure you’re the first to know about our new release, Heroku Postgres 2.0. This new release continues to build on many of the great features you may already take advantage of as well as a price drop across certain plans. This new release highlights:

Built-in alerts around what you should do for your database, bringing a new level of operational expertise built right into the product. The ability to rollback your database to a specific point in time, similar to how you can rollback bad deploys on Heroku New tiers, with our premium and enterprise tiers including high availability with automatic failover You can read more details about the features included within this release on our blog post.

You will be able to continue using your existing plans. While they will not appear in the addons:list you will be able to visit the legacy pricing page and they will continue working. We will notify you before any changes do occur to existing plans.

Cheers, The Heroku Postgres team """

Product manager of Heroku Postgres here, we did indeed lower those limits, but less for the cost and more that we saw many users hitting issues in going beyond that. Our limits exist primarily to prevent people from shooting themselves in the foot and bringing their databases down. Would be happy to hear if you're above one of those limits today and not experiencing problems - postgres@heroku.com
I have a production app that generates about 500MB of data per day. A small amount of that data needs to be kept hot in cache -- but the rest is accessed infrequently by comparison (but still in a manner that is aided by SQL). No problems at all -- system runs smooth as silk.

The problem is the application was developed based on the specs advertised (as I'm sure countless others also were). Now, the smallest database package with the same storage and connection resources is the "Standard Baku" at $2,000/month. These resources used to be available starting at $50/month. Ouch!! And even then, it is still only the "Standard" feature set -- so none of the new whiz-bang Heroku 2.0 features.

I was also disappointed in the statement of price drops - that marketing-statement of price drops alone makes me look around for alternatives. And the statement below, "Our limits exist primarily to prevent people from shooting themselves in the foot" - well that is just arrogant. If I aim at my feet, let me be a fool and pull the trigger. It is nice to see that Amazon now offers Postgresql: http://aws.typepad.com/aws/2013/11/amazon-rds-for-postgresql...

The old plans with 1TB and 500 connections made me feel comfortably ignorant of our needs, but with the new plans I must contemplate connection limits and storage capacity, and I really like not thinking about that at all.

With that said, the new features with HA and rollback is a needed addition. But I would be more likely to use HA if I could just create a follower and for an extra $50 make it automatic failover.

I just had a look at the updated pricing.

If my memory serves correctly, the same limits for storage and # of connections were in place before for the non-dev databases (starting with 50$). I don't really care for the connections, but a 64 GB storage limit can quickly become a problem for some apps, while 400MB of RAM is perfectly fine to serve the hot data...

Edit: To put this into perspective: If an app would generate 400MB of data per day (the last 24 hours would fit into the cache completely!), the app would run out of storage after 160 days...