Hdf5 as a storage for ontology instances

We are developing a platform solution that will store a lot of data with different format, origins and characteristics. The problem we are facing is: - how to best structure this data. A partner is bringing some expertise on the web semantics, and has defined an ontology - how to best store, retrieve and process a huge amount of data in a fast way; from my readings, hdf5 seems to be a great solution. My concern is how to (best) apply the ontology data model to hdf5. Does anyone has some experience in trying to implement it? A question would be, of course, how much effort does it take? Thanks