Can anybody point me to a document that explains the effect a block size has on an RMAN backup? It seems as though my databases with a smaller db_block_size backup about 20% slower than those with a bigger db_block_size.
Sorry..cannot point to any doc but are you taking a full level (0) backup with RMAN ? If so, then remember RMAN will ignore the unused blocks. Since chances of blocks being used are more if its smaller they may take more time.
Originally posted by vinit Since chances of blocks being used are more if its smaller
How so? In theory, if I have 1M of data in a 2k db_block_size, I would be using 512 blocks. If that same data uses 16K db_block_size, I would only be using 64 blocks.
However, since each database block is not 100% full with data, I am thinking that the 20% difference comes from having to read more overhead in the smaller blocks. I would like to find some documentation that supports or disproves my theory...
Originally posted by marist89
However, since each database block is not 100% full with data, I am thinking that the 20% difference comes from having to read more overhead in the smaller blocks. I would like to find some documentation that supports or disproves my theory...
You are right about the block overhead and thats what I had pointed out. Smaller the block size, more blocks used and more time for backup. Secondly it doesn't matter if the block is 100% used or not. Even if the block is empty during the backup and was used by Oracle previously it will be backed up.
Reply With Quote
Bookmarks