Using HDF5 to store ontology instances

Dear semantic web people,

First of all, I apologize if my question is unclear as I am not an expert in the semantic web. I am a beginner in the field, and to be more precise, the context of my question is the following one: we plan to develop a platform that can store massive heterogeneous data with different characteristics: dynamic data captured from sensors (which implies massive data, and time performance for storage), and  static data (from huge files, dbs, etc...).
We already planned to use an ontology to organize these data and link them. The expertise is brought to the project by a partner who already designed the ontology/data model.
The main problem we have right now is: how to store data instances considering the constraints associated to the dynamic data? From my understanding, the storage system must respect the data schema underlying the ontology. I tried to explore different solutions from Relation DBs to Graph-Oriented DBs or RDF storage systems, but none seems to suit our constraints.
I just got into HDF5 which is apparently great storage for our purpose as it is flexible, highly scalable and performant. The schema is roughly a Directed Acyclic Graph.
I was wondering if some of you have some expertise in trying to achieve what we are trying to do, if you have feedback from similar projects, or anything that could help us evaluating such a solution. All questions are welcome, and do not hesitate in criticizing this solution, of course.
Finally, if this is not the right place to ask such a question, please let me know.

Best regards,

Pierre