Prevalence Layer

Justin blogged about Bamboo.Prevalence. It's a new concept for me but I like what I see:

"Basically, Bamboo.Prevalence aims to provide transparent object persistence and synchronization to deterministic systems targeting the CLR (Common Language Runtime). No relational databases. No object-to-relational mapping goo. No SQL. Just you and your objects, isn't life great? "

Currently I'm working on a system that needs to track the not-so-simple state of a business process. From the start I was thinking in terms of ER diagrams and relationships but as I learn more about prevalence systems the more I see how this might make things quite a bit simpler.

Let's imagine you have a set of business objects who's state is easly stored. With these objects you build a business layer and on a regular interval you take a snap shot of those objects. In addition to the snapshots you log every command to the objects. If the system goes down, to bring it back you just refer to the latest snapshot and apply the changes since. That's essentially what a prevalence system does. Cool thing is all the state is stored in memory...no messy SQL involved.