When using Oracle BasicFile and SecureFile one property of the implementation is that each LOB occupies multiples of the chunk size (and in case of SecureFile the smallest chunk size is the block size).
This does luckily not happen for data which fits inline in the row, however for that data I could use a normal type. So my concern is more with data stored in the lobsegments - there the overhead can be up to a whole database block (per row) if the LOBs have unlucky sizes.
Does this change when compression is turned on?
I mean if I have a blob which is 7k and compresses to 5k It would still occupy 8k (if this is the DB blocksize). It would be somewhat unfortunate to have a tablespace with extra small blocks for (extra large) LOB segments. (But I understand that SecureFile with dynamic chunk sizes deals much better with this)
