I am assessing the possibility of my organization (software as a service org) migrating parts of our existing disk-based DBs to an in-memory cloud-based DB model (hosted in our datacenters).
The smoothest path would be to find a cloud DB that could execute dynamic SQL but I'm finding that the cloud DBs are fundamentally object-model based (i.e. no SQL).
Does anyone have experience migrating parts of an RDBMS to a cloud-based DB? If so what path did you take?
On
As for cloud based DB, you might get the ideas from what AWS is offering: http://aws.amazon.com/running_databases/
They offer three types: MySQL style web service; standard DB running in an EC2 instance; or the AWS SimpleDB, as you call object based.
If you need SQL, the previous two might be your selection. What you need to do is implement it in your private cloud.
I have wound up realizing that what I'm looking for just doesn't seem to exist in the market (a grid-DB that will scale horizontally that supports SQL).
I wound up focusing my investigation on Hazelcast and Cassandra (I prefer open source) but they are both too divergent from our existing RDBMS model to be usable without significant architectural re-work of our existing application.
The next step I have decided on is to look at pure in-memory relational DBs and maybe put one on a server with a whack of memory. Current thought is to use it for data that is accessed a lot with queries that aren't overly performant (maybe bulk load data into it nightly or something like that).
Not ideal but might be a smooth enough migration path for a high-availability production system.