This is probably something for the TerraVista peer-to-peer forum, but I'll try to answer you.
The CDB format should be much larger. It is storing the data as source, ready to be rendered by any RTP (run time publisher), not just for your VP or use case. As far as parameters, I would just use a single lod in TerraVista (it is effectively ignored, since the CDB tiles and lods are independent of what you put into TerraVista).
As to the VP startup, I can't answer that one. We don't use VP.
I would encourage you to contact Presagis product support to get expert help on Terra Vista and Vega Prime for your particular situation.
Regarding the startup time, it's not unexpected that the CDB takes longer to start up because it is stored in a client-independent, source data format. As a result, the data must be converted in real time to the internal, optimized representation of the client. This adds a small delay to the startup but the benefit is that any client can use the CDB directly.
A MetaFlight (mft) database is compiled to a binary, optimized format that is loaded virtually unchanged into the client's memory, thus loading faster. The downside is that this database can only be used by Vega Prime and cannot be directly shared with other clients, like the CDB can.
I am in the process of generating a very high quality-large area terrain through Terra-Vista. Currently, I am generating .mft and cdb as my output types to cross check optimum rendering in Vega-Prime. Need expert opinions on following points:
1) what should be the Project Parameters for cdb output in TV? I mean, how many LODs do i need to fix (if my observer never comes down below 20km!)
2) CDBOutput is taking huge disk space as compared to .mft!!
3) There is a rendering delay for CDBOutput in VP whereas .mft renders instantly
any suggestion would be greatly appreciated