[sjdev]

We have been having that discussion internally at the company I work for. After the initial shock wore off and we started deciding what we would do with the projects currently on Parse, we realized that this could very well be a good thing. Before this, we were not able to use Parse for larger clients or clients with specific database hosting needs, but that may no longer be an issue. I'm excited to see what the future holds for Open Source Parse.

[delphinius81]

We spent a few days investigating alternatives and then decided to just wait and see what happens over the next few months. We want to switch to DynamoDB instead of MongoDB due to cheaper hosting costs through AWS for the amount of data we use. There's already someone working on cleaning up the DB code to allow for switching DB providers.

[pritambarhate]

In my study of Dynamo DB I found that it might actually be more expensive than managed Mongo DB. But I didn't get a chance to test this out because of lack of budget as we decided to build against Mongo DB rather than build for both and check which one is more expensive.

The problem with Dynamo DB is that one needs to provision the read and write capacity in advance, which means that one needs to over provision in order to be sure. Also even for things like exporting database which requires lots of reads in short time, one needs to provision the read capacity. This can make the back ups very expensive. However I am basing this from my study of their docs as we were building our application. (A chat server).

It will be good if somebody can share their actual experience about Dynamo DB especially from the perspective of cost. Yes, almost unlimited scalability is very attractive, but I found that I had to think about data access patterns a lot to ensure that we will stay in acceptable read write quota for Dynamo DB. This was causing a lot of delay in our MVP. So we decided to go with Mongo DB which we knew we could scale fairly beyond our MVP and may be rewrite the data access layer later on if we achieve a mass that requires a switch to something like Dynamo Db or manual sharding.