Hi all,

I am starting a new large client server based project and I am deciding what system our team should use for persistence.

I am very keen to either use an OODB, ORDBMS or a OR Mapper. Nhibernate looks to be one of the most impressive mappers on the market for .net although I do worry that it is in beta. Firstly are there any bugs remaining in this product that I should stress about?

I am going to be deploying our application layer into a scale out architecture. The plan is to come up with a cool system so we can add and remove servers as required with no down time.

I also want to be able to add servers as required to the persistence layer so therefore I am looking at database clustering.

How does hibernate work with regards to clusters of database servers? What happens if two application servers are accessing the same object and perform a flush concurrently or if one overwrites the other?

I take it there must be in built support for this. Or should I steer away from OR mapping with large projects? I am just keen to avoid writing SQL code and cluttering up my nice object model ;) And I am yet to find a suitable OODB for .net - although I am keen to hear any suggestions.

Your thoughts are welcomed.

Kind Regards,
Mark :shock:

OR mappers are usually better than datasets for large complicated systems. But expect to have problems with binding objects to windows forms, it's not that easy as with datasets. NHibernate supports both forms of concurrency - optimistic (timestamps) and pessimistic (locking). It doesn't support multiple databases well - you can use several databases, but you can't have associations across databases.

As for bugs, there are some (see http://jira.nhibernate.org/), we're going to fix most of them for the 0.9 release. NHibernate is already being used by many people in quite large applications.

About OO databases - the only free one I have heard of is db4o, and I don't know how well it scales. There's also an OO database by Versant, but it's quite expensive.