In general RDBMS hold gridded data as Binary large Objects, short BLOBS.
An OPeNDAP servers hosts BLOBS. A system with an RDBMS as core can use OPeNDAP data in three ways, or as a combi of those:

  1. copy binary data into RDBMS as BLOB. This option is best suited for recent not-too-large data.
  2. copy netCDF files to a local LAN file server, and add a link to the RDBMS, i.e. make a local cache.
  3. add a link to the OPeNDAP url, and only access the data when you rally need them. Why copy data, when they are online available, You don;t copy YouTube moviers either do you? This option is best suited for old very-large data. This combi between an RDBMS that links to remote OPeNDAP data is the basis for a data cloud. A pilot data cloud is currently being explored in the Dutch NMDC.