ahh, interesting. I didn't know about the "minimum extent" clause.

Do you think that an optimal minimum extent size should be db_block_size * db_multiblock_read_count?