Elasticity without inconsistency: Atomikos and the LogCloud

So you want elastic scaling. What about transaction recovery?

It's the cloud era, and you want to build elastic applications with distributed transactions. You would typically need a local disk to persist transaction logs and perform recovery. This would then defeat the purpose of elastic scaling since by definition nodes can disappear. In addition, the cloud platform where you are deploying to might not even support disk access in the first place. So you would have to make this tough choice: either you get elasticity, or consistency but not both.

We now have a solution and it's called The LogCloud: a dedicated logging and recovery node that allows you to add / remove processes dynamically in your application cluster. Transaction processing will never be the same since you can now have elasticity without inconsistency!